Efficient retrieval of external knowledge bases and web pages is crucial for enhancing the reasoning abilities of LLMs. Previous works on training LLMs to leverage external retrievers for solving complex problems have predominantly employed end-to-end reinforcement learning. However, these approaches neglect supervision over the reasoning process, making it difficult to guarantee logical coherence and rigor. To address these limitations, we propose Thinker, a hierarchical thinking model for deep search through multi-turn interaction, making the reasoning process supervisable and verifiable. It decomposes complex problems into independently solvable sub-problems, each dually represented in both natural language and an equivalent logical function to support knowledge base and web searches. Concurrently, dependencies between sub-problems are passed as parameters via these logical functions, enhancing the logical coherence of the problem-solving process. To avoid unnecessary external searches, we perform knowledge boundary determination to check if a sub-problem is within the LLM's intrinsic knowledge, allowing it to answer directly. Experimental results indicate that with as few as several hundred training samples, the performance of Thinker is competitive with established baselines. Furthermore, when scaled to the full training set, Thinker significantly outperforms these methods across various datasets and model sizes.

Xinyu Tang, Yuliang Zhan, Zhixun Li, Xin Zhao, Zhenduo Zhang, Zujie Wen, Zhiqiang Zhang, Jun Zhou. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Jujia Zhao, Zhaoxin Huan, Zihan Wang, Xiaolu Zhang, Jun Zhou, Suzan Verberne, Zhaochun Ren. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Fengqi Zhu, Rongzhen Wang, Shen Nie, Xiaolu Zhang, Chunwei Wu, Jun Zhou, Yankai Lin, Ji-Rong Wen, Chongxuan Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning. The method introduces two simple strategies that modify training data by previewing partial answer prefixes, thereby minimizing perturbations to the model's initial token distributions and maintaining its built-in safety mechanisms. Comprehensive experiments demonstrate that LookAhead Tuning effectively maintains model safety without sacrificing robust performance on downstream tasks. Our findings position LookAhead Tuning https://github.com/zjunlp/LookAheadTuning as a reliable and efficient solution for the safe and effective adaptation of LLMs.

Large Language Models (LLMs) have demonstrated strong potential for generative recommendation by leveraging rich semantic knowledge. However, existing LLM-based recommender systems struggle to effectively incorporate collaborative filtering (CF) signals, due to a fundamental mismatch between item-level preference modeling in CF and token-level next-token prediction (NTP) optimization in LLMs. Prior approaches typically treat CF as contextual hints or representation bias, and resort to multi-stage training to reduce behavioral–semantic space discrepancies, leaving CF unable to explicitly regulate LLM generation. In this work, we propose Token-level Collaborative Alignment for Recommendation (TCA4Rec), a model-agnostic and plug-and-play framework that establishes an explicit optimization-level interface between CF supervision and LLM generation. TCA4Rec consists of (i) Collaborative Tokenizer, which projects raw item-level CF logits into token-level distributions aligned with the LLM token space, and (ii) Soft Label Alignment, which integrates these CF-informed distributions with one-hot supervision to optimize a soft NTP objective. This design preserves the generative nature of LLM training while enabling collaborative alignment with essential user preference of CF models. We highlight TCA4Rec is compatible with arbitrary traditional CF models and generalizes across a wide range of decoder-based LLM recommender architectures. Moreover, it provides an explicit mechanism to balance behavioral alignment and semantic fluency, yielding generative recommendations that are both accurate and controllable. Extensive experiments demonstrate that TCA4Rec consistently improves recommendation performance across a broad spectrum of CF models and LLM-based recommender systems. Our code is available at https://github.com/critical88/TCA4Rec

Implicit feedback is central to modern recommender systems but is inherently noisy, often impairing model training and degrading user experience. At scale, such noise can mislead learning processes, reducing both recommendation accuracy and platform value. Existing denoising strategies typically overlook the entity-specific nature of noise while introducing high computational costs and complex hyperparameter tuning. To address these challenges, we propose EARD (Entity-Aware Reliability-Driven Denoising), a lightweight framework that shifts the focus from interaction-level signals to entity-level reliability. Motivated by the empirical observation that training loss correlates with noise, EARD quantifies user and item reliability via their average training losses as a proxy for reputation, and integrates these entity-level factors with interaction-level confidence. The framework is model-agnostic, computationally efficient, and requires only two intuitive hyperparameters. Extensive experiments across multiple datasets and backbone models demonstrate that EARD yields substantial improvements over state-of-the-art baselines (e.g., up to 27.01% gain in NDCG@50), while incurring negligible additional computational cost. Comprehensive ablation studies and mechanism analyses further confirm EARD's robustness to hyperparameter choices and its practical scalability. These results highlight the importance of entity-aware reliability modeling for denoising implicit feedback and pave the way for more robust recommendation research.

Multi-Source Domain Adaptation (MSDA) is receiving increased focus as a technique for reliably transferring knowledge from several source domains to a specific target domain. Traditional approaches generally operate under the assumption that samples in each source conform to a consistent distribution. However, real-world conditions often involve samples derived from diverse distributions, as well as potential data imbalance among different domains. Addressing these challenges, we introduce a novel and trustworthy framework, Multi-source Reconstructed Domain Adaptation (MSRDA), designed to enhance adaptation efficacy while maintaining robust performance and reliability across heterogeneous data sources. To start with,we delve into the latent mixed distributions of each source using clustering techniques, followed by the reconstruction of the latent domains following the original distribution. Additionally, we introduce an adaptive weighting mechanism to mitigate data imbalances.In cases where an latent domain consists of only a few samples, the global features are identified and dominate in that particular domain to help avoid overfitting. Moreover, given the difficulties of optimizing clustering while updating the model,we apply ExpectationMaximization (EM) algorithm to iteratively perform domain reconstruction and domain adaptation.Experiments are performed on two public datasets and one real-world datasets, and experimental results demonstrate that our MSRDA can effectively achieve multi-source domain adaptation through re-. constructing source domain with identified latent domains.

This is a corrigendum for the article “One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems” published in ACM Trans. Inf. Syst . 43, 5, Article 118 (July 2025), 27 pages.

Noisy labels are both inevitable and problematic in machine learning methods, as they negatively impact models' generalization ability by causing overfitting. In the context of learning with noise, the transition matrix plays a crucial role in the design of statistically consistent algorithms. However, the transition matrix is often considered unidentifiable. One strand of methods typically addresses this problem by assuming that the transition matrix is instance-independent; that is, the probability of mislabeling a particular instance is not influenced by its characteristics or attributes. This assumption is clearly invalid in complex real-world scenarios. To better understand the transition relationship and relax this assumption, we propose to study the data generation process of noisy labels from a causal perspective. We discover that an unobservable latent variable can affect either the instance itself, the label annotation procedure, or both, which complicates the identification of the transition matrix. To address various scenarios, we have unified these observations within a new causal graph. In this graph, the input instance is divided into a noise-resistant component and a noise-sensitive component based on whether they are affected by the latent variable. These two components contribute to identifying the “causal transition matrix”, which approximates the true transition matrix with theoretical guarantee. In line with this, we have designed a novel training framework that explicitly models this causal relationship and, as a result, achieves a more accurate model for inferring the clean label.

Complex question answering (QA) is a challenging task in artificial intelligence research which requires reasoning based on related knowledge. The retrieval-augmented generation (RAG) based on large language models (LLMs) have become one promising solution in QA. To facilitate RAG more effectively, the LLM needs to precisely evaluate knowledge required in QA. That is, first, the LLM needs to examine its knowledge boundary (what the LLM does not know) to retrieve external knowledge as supplement. Second, the LLM needs to evaluate the utility of the retrieved knowledge (whether it helps in reasoning) for robust RAG. To this end, in this paper, we propose a novel Question Answering with Knowledge Evaluation (KEQA) framework to promote the effectiveness and efficiency of RAG in QA. First, inspired by quizzes in classroom, we propose a quiz-based method to precisely examine the knowledge state of the uninterpretable LLM for QA. We ask indicative quizzes on each required knowledge, and inspect whether the LLM can consistently answer the quiz to examine its knowledge boundary. Second, we retrieve the unknown knowledge from external source, and evaluate its utility to pick the helpful ones for reasoning. We design a reasoning-based metric to evaluate utility, and construct a demonstration set in training data for reference to guide knowledge picking in inference. We conduct extensive experiments on four widely-used QA datasets, and the results demonstrate the effectiveness of the proposed method.

The safety mechanisms of large language models (LLMs) exhibit notable fragility, as even fine-tuning on datasets without harmful content may still undermine their safety capabilities.Meanwhile, existing safety alignment methods predominantly rely on the fine-tuning process, which inadvertently leads to the increased complexity and computational resources required.To address these issues, we introduce LSSF, a novel safety re-alignment framework with Low-Rank Safety Subspace Fusison.Our proposed method exploits the low-rank characteristics of safety information in LLMs by constructing a low-rank projection matrix to extract the principal components of safety vectors.Notably, this projection matrix represents the low-rank safety subspace of the LLMs, which we have observed to remain stable during fine-tuning process and is isolated from the model's general capabilities.These principal components are used to effectively restore safety alignment when combined with fine-tuned LLMs through linear arithmetic.Additionally, to account for the varying encoding densities of safety information across different layers of LLMs, we propose a novel metric called safety singular value entropy.This metric quantifies the encoding density and allows for the dynamic computation of the safety-critical rank for each safety vector.Extensive experiments demonstrate that our proposed post-hoc alignment method can effectively restore the safety alignment of finetuned models with minimal impact on their performance in downstream tasks.

Enhancing the mathematical reasoning of large language models (LLMs) demands high-quality training data, yet conventional methods face critical challenges in scalability, cost, and data reliability. To address these limitations, we propose a novel program-assisted synthesis framework that systematically generates a high-quality mathematical corpus with guaranteed diversity, complexity, and correctness. This framework integrates mathematical knowledge systems and domain-specific tools to create executable programs. These programs are then translated into natural language problem-solution pairs and vetted by a bilateral validation mechanism that verifies solution correctness against program outputs and ensures program-problem consistency. We have generated 12.3 million such problem-solving triples. Experiments demonstrate that models fine-tuned on our data significantly improve their inference capabilities, achieving state-of-the-art performance on several benchmark datasets and showcasing the effectiveness of our synthesis approach.

Reinforcement learning from human feedback (RLHF) offers a promising approach to aligning large language models (LLMs) with human preferences.Typically, a reward model is trained or supplied to act as a proxy for humans in evaluating generated responses during the reinforcement training phase.However, current reward models operate as sequence-to-one models, allocating a single, sparse, and delayed reward to an entire output sequence.This approach may overlook the significant contributions of individual tokens toward the desired outcome.To this end, we propose a more finegrained, token-level guidance approach for RL training.Specifically, we introduce RED, a novel REward reDistribition method that evaluates and assigns specific credit to each token using an off-the-shelf reward model.Utilizing these fine-grained rewards enhances the model's understanding of language nuances, leading to more precise performance improvements.Notably, our method does not require modifying the reward model or introducing additional training steps, thereby incurring minimal computational costs.Experimental results across diverse datasets and tasks demonstrate the superiority of our approach.

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet they often refuse to answer legitimate queries-a phenomenon known as overrefusal.Overrefusal typically stems from over-conservative safety alignment, causing models to treat many reasonable prompts as potentially risky.To systematically understand this issue, we probe and leverage the models' safety decision boundaries to analyze and mitigate overrefusal.Our findings reveal that overrefusal is closely tied to misalignment at these boundary regions, where models struggle to distinguish subtle differences between benign and harmful content.Building on these insights, we present RASS, an automated framework for prompt generation and selection that strategically targets overrefusal prompts near the safety boundary.By harnessing steering vectors in the representation space, RASS efficiently identifies and curates boundary-aligned prompts, enabling more effective and targeted mitigation of overrefusal.This approach not only provides a more precise and interpretable view of model safety decisions but also seamlessly extends to multilingual scenarios.We have explored the safety decision boundaries of various LLMs and construct the MORBENCH evaluation set to facilitate robust assessment of model safety and helpfulness across multiple languages.

The Direct Segment Anything Model (DirectSAM) excels in class-agnostic contour extraction. In this paper, we explore its use by applying it to optical remote sensing imagery, where semantic contour extraction—such as identifying buildings, road networks, and coastlines-holds significant practical value. Those applications are currently handled via training specialized small models separately on small datasets in each domain. We introduce a foundation model derived from DirectSAM, termed DirectSAM-RS, which not only inherits the strong segmentation capability acquired from natural images, but also benefits from a large-scale dataset we created for remote sensing semantic contour extraction. This dataset comprises over 34k image-text-contour triplets, making it at least 30 times larger than individual dataset. DirectSAM-RS integrates a prompter module: a text encoder and cross-attention layers attached to the DirectSAM architecture, which allows flexible conditioning on target class labels or referring expressions. We evaluate the DirectSAM-RS in both zero-shot and fine-tuning setting, and demonstrate that it achieves state-of-the-art performance across several downstream benchmarks.

In this paper, we seek to carefully prompt a Large Language Model (LLM) with domain-level knowledge as a better marketing-oriented knowledge miner for marketing-oriented knowledge graph construction, which is non-trivial, suffering from several inevitable issues in real-world marketing scenarios, i.e., uncontrollable relation generation of LLMs, insufficient prompting ability of a single prompt, unaffordable deployment cost of LLMs. To this end, we propose PAIR, a novel Progressive prompting Augmented mIning fRamework for harvesting marketing-oriented knowledge graph with LLMs. In particular, we reduce the pure relation generation to an LLM-based adaptive relation filtering process through knowledge-empowered prompting technique. Next, we steer LLMs for entity expansion with progressive prompting augmentation, followed by a reliable aggregation with comprehensive consideration of both self-consistency and semantic relatedness. In terms of online serving, we specialize in a small and white-box PAIR (i.e., LightPAIR), which is fine-tuned with a high-quality corpus provided by a strong teacher-LLM. Extensive experiments and practical applications in audience targeting verify the effectiveness of the proposed (Light)PAIR.

Zero-shot composed image retrieval (ZS-CIR) is a rapidly growing area with significant practical applications, allowing users to retrieve a target image by providing a reference image and a relative caption describing the desired modifications. Existing ZS-CIR methods often struggle to capture fine-grained changes and integrate visual and semantic information effectively. They primarily rely on either transforming the multimodal query into a single text using image-to-text models or employing large language models for target image description generation, approaches that often fail to capture complementary visual information and complete semantic context. To address these limitations, we propose a novel Fine-Grained Zero-Shot Composed Image Retrieval method with Complementary VisualSemantic Integration (CVSI). Specifically, CVSI leverages three key components: (1) Visual Information Extraction, which not only extracts global image features but also uses a pre-trained mapping network to convert the image into a pseudo token, combining it with the modification text and the objects most likely to be added. (2) Semantic Information Extraction, which involves using a pre-trained captioning model to generate multiple captions for the reference image, followed by leveraging an LLM to generate the modified captions and the objects most likely to be added. (3) Complementary Information Retrieval, which integrates information extracted from both the query and database images to retrieve the target image, enabling the system to efficiently handle retrieval queries in a variety of situations. Extensive experiments on three public datasets (e.g., CIRR, CIRCO, and FashionIQ) demonstrate that CVSI significantly outperforms existing state-of-the-art methods. Our code is available at https://github.com/yyc6631/CVSI.

Catastrophic forgetting (CF) poses a significant challenge in machine learning, where a model forgets previously learned information upon learning new tasks. Despite the advanced capabilities of Large Language Models (LLMs), they continue to face challenges with CF during continual learning. The majority of existing research focuses on analyzing forgetting patterns through a singular training sequence, thereby overlooking the intricate effects that diverse tasks have on model behavior. Our study explores CF across various settings, discovering that model forgetting is influenced by both the specific training tasks and the models themselves. To this end, we interpret forgetting by examining the function vector (FV), a compact representation of functions in LLMs, offering a model-dependent indicator for the occurrence of CF. Through theoretical and empirical analyses, we demonstrated that CF in LLMs primarily stems from biases in function activation rather than the overwriting of task processing functions. Leveraging these insights, we propose a novel function vector guided training methodology, incorporating a regularization technique to stabilize the FV and mitigate forgetting. Empirical tests on four benchmarks confirm the effectiveness of our proposed training method, substantiating our theoretical framework concerning CF and model function dynamics.

High-quality data plays a critical role in the pretraining and fine-tuning of large language models (LLMs), even determining their performance ceiling to some degree. Consequently, numerous data selection methods have been proposed to identify subsets of data that can effectively and efficiently enhance model performance. However, most of these methods focus on general data selection and tend to overlook the specific nuances of domain-related data. In this paper, we introduce MASS, a Mathematical data Selection framework using the Skill graph for pretraining LLMs in the mathematical reasoning domain. By taking into account the unique characteristics of mathematics and reasoning, we construct a skill graph that captures the mathematical skills and their interrelations from a reference dataset. This skill graph guides us in assigning quality scores to the target dataset, enabling us to select the top-ranked subset which is further used to pretrain LLMs. Experimental results demonstrate the efficiency and effectiveness of MASS across different model sizes (1B and 7B) and pretraining datasets (web data and synthetic data). Specifically, in terms of efficiency, models trained on subsets selected by MASS can achieve similar performance to models trained on the original datasets, with a significant reduction in the number of trained tokens - ranging from 50% to 70% fewer tokens. In terms of effectiveness, when trained on the same amount of tokens, models trained on the data selected by MASS outperform those trained on the original datasets by 3.3% to 5.9%. These results underscore the potential of MASS to improve both the efficiency and effectiveness of pretraining LLMs.

Although masked image generation models and masked diffusion models are designed with different motivations and objectives, we observe that they can be unified within a single framework. Building upon this insight, we carefully explore the design space of training and sampling, identifying key factors that contribute to both performance and efficiency. Based on the improvements observed during this exploration, we develop our model, referred to as \textbf{eMIGM}. Empirically, eMIGM demonstrates strong performance on ImageNet generation, as measured by Fréchet Inception Distance (FID). In particular, on ImageNet $256\times256$, with similar number of function evaluations (NFEs) and model parameters, eMIGM outperforms the seminal VAR. Moreover, as NFE and model parameters increase, eMIGM achieves performance comparable to the state-of-the-art continuous diffusion model REPA while requiring less than 45\% of the NFE. Additionally, on ImageNet $512\times512$, eMIGM outperforms the strong continuous diffusion model EDM2. Code is available at \url{https://github.com/ML-GSAI/eMIGM}.

Uplift modeling is crucial for identifying individuals likely to respond to a treatment in applications like marketing and customer retention, but evaluating these models is challenging due to the inaccessibility of counterfactual outcomes in real-world settings. In this paper, we identify a fundamental limitation in existing evaluation metrics, such as the uplift and Qini curves, which fail to rank individuals with binary negative outcomes accurately. This can lead to biased evaluations, where biased models receive higher curve values than unbiased ones, resulting in suboptimal model selection. To address this, we propose the Principled Uplift Curve (PUC), a novel evaluation metric that assigns equal curve values of individuals with both positive and negative binary outcomes, offering a more balanced and unbiased assessment. We then derive the Principled Uplift Loss (PUL) function from the PUC and integrate it into a new uplift model, the Principled Treatment and Outcome Network (PTONet), to reduce bias during uplift model training. Experiments on both simulated and real-world datasets demonstrate that the PUC provides less biased evaluations, while PTONet outperforms existing methods. The source code is available at: https://github.com/euzmin/PUC.

Recently, ChatGPT or InstructGPT like large language models (LLM) has made a significant impact in the AI world. Many works have attempted to reproduce the complex InstructGPT's training pipeline, namely Reinforcement Learning with Human Feedback (RLHF). However, the mainstream distributed RLHF training methods typically adopt a fixed model placement strategy, referred to as the All-colocated strategy. This strategy treats all four interdependent models involved in RLHF as a single entity, distributing them across all devices and applying parallelism techniques designed for a single model, regardless of the workload heterogeneity inherent to each model. As a result, this strategy exacerbates memory constraints and generation bottlenecks in RLHF training. To address these issues, we propose a flexible model placement framework that offers two general and agile model placement strategies. The Interleaving strategy helps improve memory utilization and communication efficiency of RLHF training by placing models without dependencies on exclusive devices with careful orchestration. On the other hand, the Disaggregated strategy improves the throughput of model training by separating the training and inference runtime of the RLHF pipeline with additional shadow models. Furthermore, our framework provides a simple user interface and guidelines to easily and flexibly configure these strategies in various training scenarios. Our experiments have shown that our strategy can achieve notable improvements up to$12\times$, compared to Deepspeed-Chat or trlX. The results highlight the effectiveness and adaptability of our methods in accelerating the training of distributed RLHF.

The rapid growth of health insurance and the rising incidence of fraudulent claims underscore the necessity for an efficient and professional claims management system. However, there is a noticeable lack of shared relevant experience from previous research in this field. In response to this challenge, we introduce AntAkso, a robust claims management system specifically designed for health insurance operations within Alipay. AntAkso incorporates a digital and professional management system, achieving a notable decrease in the volume of false claims, reduction in administrative costs, and heightened satisfaction among its policyholders. We begin by highlighting the core components of this system, including the case stratification, hospital recommendation, and case dispatch modules, along with the pivotal algorithms employed, i.e., the fraud detection, recommendation, and robust satisficing algorithms. We also detail the system's implementation and deployment. We substantiate the proposed system's effectiveness and efficiency with empirical evidence from experiments on a large set of real-world health insurance claims data.

In recent years, Graph Neural Networks (GNNs) become very effective methods to utilize graphs and have been applied to many real-world applications, including recommendation, advertisement, and financial fraud detection. In fact, GNNs are mostly trained and test in the environments with the same distribution. However, in the real cases, selection bias are inevitably existed in both the node features and the graph structures, which will lead to serious impact on the GNN performance. Several works of literature have investigated the out-of-distribution (OOD) problem on the feature distribution, but little research specifically studies the effect caused by the bias of graph structure. However, graph structure is very fundamental for GNNs since it greatly affects the message propagation mechanism.

The capabilities of large language models (LLMs) are widely regarded as relying on autoregressive models (ARMs). We challenge this notion by introducing *LLaDA*, a diffusion model trained from scratch under the pre-training and supervised fine-tuning (SFT) paradigm. LLaDA employs a forward data masking process and a reverse generation process, parameterized by a Transformer to predict masked tokens. It provides a principled generative approach for probabilistic inference by optimizing a likelihood lower bound. Across extensive benchmarks on general tasks, math, code, and so on, LLaDA demonstrates strong *scalability* and performs comparably to our self-constructed ARM baselines. Remarkably, LLaDA 8B is competitive with strong LLMs like LLaMA3 8B in *in-context learning* and, after SFT, exhibits impressive *instruction-following* abilities in case studies such as multi-turn dialogue. Moreover, LLaDA addresses the reversal curse, surpassing GPT-4o in a reversal poem completion task. Our findings show the promise of diffusion models for language modeling at scale and challenge the common assumption that core LLM capabilities discussed above inherently depend on ARMs. Project page and codes: \url{https://ml-gsai.github.io/LLaDA-demo/}.

We propose a novel AutoRegressive Generation-based paradigm for image Segmentation (ARGenSeg), achieving multimodal understanding and pixel-level perception within a unified framework. Prior works integrating image segmentation into multimodal large language models (MLLMs) typically employ either boundary points representation or dedicated segmentation heads. These methods rely on discrete representations or semantic prompts fed into task-specific decoders, which limits the ability of the MLLM to capture fine-grained visual details. To address these challenges, we introduce a segmentation framework for MLLM based on image generation, which naturally produces dense masks for target objects. We leverage MLLM to output visual tokens and detokenize them into images using an universal VQ-VAE, making the segmentation fully dependent on the pixel-level understanding of the MLLM. To reduce inference latency, we employ a next-scale-prediction strategy to generate required visual tokens in parallel. Extensive experiments demonstrate that our method surpasses prior state-of-the-art approaches on multiple segmentation datasets with a remarkable boost in inference speed, while maintaining strong understanding capabilities.

Achieving balanced alignment of large language models (LLMs) in terms of Helpfulness, Honesty, and Harmlessness (3H optimization) constitutes a cornerstone of responsible AI. Existing methods like data mixture strategies face limitations, including heavy reliance on expert knowledge and conflicting optimization signals. While model merging offers parameter-level conflict-resolution strategies through integrating specialized models' parameters, its potential for 3H optimization remains underexplored. This paper systematically compares the effectiveness of model merging and data mixture methods in constructing 3H-aligned LLMs for the first time, revealing previously overlooked collaborative and conflict relationships among the 3H dimensions and discussing the advantages and drawbacks of data mixture (\textit{data-level}) and model merging (\textit{parameter-level}) methods in mitigating the conflict for balanced 3H optimization. Specially, we propose a novel \textbf{R}eweighting \textbf{E}nhanced task \textbf{S}ingular \textbf{M}erging method, \textbf{RESM}, through outlier weighting and sparsity-aware rank selection strategies to address the challenges of preference noise accumulation and layer sparsity adaptation inherent in 3H-aligned LLM merging. Extensive evaluations can verify the effectiveness and robustness of RESM compared to previous data mixture (2\%-5\% gain) and model merging (1\%-3\% gain) methods in achieving balanced LLM alignment.

Estimating individual treatment effects (ITE) from observational data is a critical task across various domains. However, many existing works on ITE estimation overlook the influence of hidden confounders, which remain unobserved at the individual unit level. To address this limitation, researchers have utilized graph neural networks to aggregate neighbors' features to capture the hidden confounders and mitigate confounding bias by minimizing the discrepancy of confounder representations between the treated and control groups. Despite the success of these approaches, practical scenarios often treat all features as confounders and involve substantial differences in feature distributions between the treated and control groups. Confusing the adjustment and confounder and enforcing strict balance on the confounder representations could potentially undermine the effectiveness of outcome prediction. To mitigate this issue, we propose a novel framework called the Graph Disentangle Causal model (GDC) to conduct ITE estimation in the network setting. GDC utilizes a causal disentangle module to separate unit features into adjustment and confounder representations. Then we design a graph aggregation module consisting of three distinct graph aggregators to obtain adjustment, confounder, and counterfactual confounder representations. Finally, a causal constraint module is employed to enforce the disentangled representations as true causal factors. The effectiveness of our proposed method is demonstrated by conducting comprehensive experiments on two networked datasets.

Graph Neural Networks (GNNs) have achieved great success in dealing with non-Euclidean graph-structured data and have been widely deployed in many real-world applications. However, their effectiveness is often jeopardized under class-imbalanced training sets. Most existing studies have analyzed class-imbalanced node classification from a supervised learning perspective, they do not fully utilize the large number of unlabeled nodes in semi-supervised scenarios. We claim that the supervised signal is just the tip of the iceberg and a large number of unlabeled nodes have not yet been effectively utilized. In this work, we propose IceBerg, a debiased self-training framework to address the class-imbalanced and few-shot challenges for GNNs at the same time. Specifically, to figure out the Matthew effect and label distribution shift in self-training, we propose Double Balancing, which can largely improve the performance of existing baselines with just a few lines of code as a simple plug-and-play module. Secondly, to enhance the long-range propagation capability of GNNs, we disentangle the propagation and transformation operations of GNNs. Therefore, the weak supervision signals can propagate more effectively to address the few-shot issue. In summary, we find that leveraging unlabeled nodes can significantly enhance the performance of GNNs in class-imbalanced and few-shot scenarios, and even small, surgical modifications can lead to substantial performance improvements. Systematic experiments on benchmark datasets show that our method can deliver considerable performance gain over existing class-imbalanced node classification baselines. Additionally, due to IceBerg's outstanding ability to leverage unsupervised signals, it also achieves state-of-the-art results in few-shot node classification scenarios. The code of IceBerg is available at: https://github.com/ZhixunLEE/IceBerg.

Due to the high cost of communication, federated learning (FL) systems need to sample a subset of clients that are involved in each round of training. As a result, client sampling plays an important role in FL systems as it affects the convergence rate of optimization algorithms used to train machine learning models. Despite its importance, there is limited work on how to sample clients effectively. In this paper, we cast client sampling as an online learning task with bandit feedback, which we solve with an online stochastic mirror descent (OSMD) algorithm designed to minimize the sampling variance. We then theoretically show how our sampling method can improve the convergence speed of federated optimization algorithms over the widely used uniform sampling. Through both simulated and real data experiments, we empirically illustrate the advantages of the proposed client sampling algorithm over uniform sampling and existing online learning-based sampling strategies. The proposed adaptive sampling procedure is applicable beyond the FL problem studied here and can be used to improve the performance of stochastic optimization procedures such as stochastic gradient descent and stochastic coordinate descent.

Semantic segmentation of remote sensing images (RSIs) plays a pivotal role in advancing geospatial analyses and applications across diverse fields, such as urban planning and environmental monitoring. Traditional learning paradigms predominantly utilize Euclidean spaces for feature extraction. This approach can introduce spatial distortions when representing objects, as Euclidean architectures typically focus on locality and are optimized for grid data, not always yielding optimal geometrical representations for data structured in non-Euclidean spaces. To address these problems, we propose EAAHNet, the first fully hyperbolic neural network designed for semantic segmentation of RSIs. EAAHNet employs the Lorentz model to reformalize conventional Euclidean-based neural network operations, ensuring the preservation of hyperbolic properties. Furthermore, to account for the inherently Euclidean nature of ground objects, we propose a Euclidean affinity-augmented hyperbolic attention module (EAAHAM) that enriches contextual dependencies through an attention fusion manner. This enhancement significantly improves the network’s capacity to discern pixel-wise semantics. Extensive experiments conducted on the ISPRS Vaihingen, ISPRS Potsdam, and LoveDA datasets demonstrate EAAHNet’s superior performance over several state-of-the-art methods. Additionally, the ablation study verifies the impacts of EAAHAM.

Graph augmentation is the key component to reveal instance-discriminative features of a graph as its rationale—an interpretation for it—in graph contrastive learning (GCL). Existing rationale-aware augmentation mechanisms in GCL frameworks roughly fall into two categories and suffer from inherent limitations: (1) non-heuristic methods with the guidance of domain knowledge to preserve salient features, which require expensive expertise and lack generality, or (2) heuristic augmentations with a co-trained auxiliary model to identify crucial substructures, which face not only the dilemma between system complexity and transformation diversitybut also the instability stemming from the co-training of two separated sub-models. Inspired by recent studies on transformers, we propose self-attentive rationale-guided GCL (SR-GCL), which integrates rationale generator and encoder together, leverages the self-attention values in transformer module as a natural guidance to delineate semantically informative substructures from both node- and edge-wise perspectives, and contrasts on rationale-aware augmented pairs. On real-world biochemistry datasets, visualization results verify the effectiveness and interpretability of self-attentive rationalization, and the performance on downstream tasks demonstrates the state-of-the-art performance of SR-GCL for graph model pre-training. Codes are available at https://github.com/lsh0520/SR-GCL .

In graph classification, the out-of-distribution (OOD) issue is attracting great attention. To address this issue, a prevailing idea is to learn stable features, on the assumption that they are substructures causally determining the label and that their relationship with the label is stable to the distributional uncertainty. In contrast, the complementary parts termed environmental features, fail to determine the label solely and hold varying relationships with the label, thus ascribed to the possible reason for the distribution shift. Existing generalization efforts mainly encourage the model’s insensitivity to environmental features. While the sensitivity to stable features is promising to distinguish the crucial clues from the distributional uncertainty but largely unexplored. A paradigm of simultaneously exploring the sensitivity to stable features and insensitivity to environmental features is until-now lacking to achieve the generalizable graph classification, to the best of our knowledge. In this work, we conjecture that generalizable models should be sensitive to stable features and insensitive to environmental features. To this end, we propose a simple yet effective augmentation strategy for graph classification: Equivariant and Invariant Cross-Data Augmentation (EI-CDA). By employing equivariance, given a pair of input graphs, we first estimate their stable and environmental features via masks. Then, we linearly mix the estimated stable features of two graphs and encourage the model predictions faithfully reflect their mixed semantics. Meanwhile, by using invariance, we swap the estimated environmental features of two graphs and keep the predictions invariant. This simple yet effective strategy endows the models with both sensitivity to stable features and insensitivity to environmental features. Extensive experiments show that EI-CDA significantly improves performance and outperforms leading baselines. Our codes are available at: https://github.com/yongduosui/EI-GNN .

The data sparsity problem has been a long-standing obstacle towards achieving better recommendation performance since it is miserable to estimate the user’s interests from limited historical behaviors. The pre-training paradigm, i.e., learning universal knowledge across a wide spectrum of domains, has increasingly become a new de-facto practice in many fields, especially for adaption to new domains. The merit of this superior generalizability renders it a natural choice to tackle the data sparsity problem for various recommendation scenarios. Hence, several efforts mainly follow masked language modeling or simple data augmentation via contrastive learning to build a pre-trained recommendation model. Our recent work (namely Miracle ) suggests that the common treatment utilizing the masked language modeling is not sufficient for pre-training a recommender system, since a user’s intent could be more complex than predicting the next word or item. The encouraging results demonstrate that the multi-interest modeling could significantly push the frontier of recommender system pre-training. Nevertheless, how to accommodate the temporal dynamics of the user interests seems to be underexplored under both single vector representation and multi-interest schemes. In this article, we aim to incorporate sophisticated temporal information modeling with the current advance in this line. More specifically, we extend Miracle by further considering relative position information and two kinds of relative time interval information jointly when performing multi-interest learning. Then, a sequential process for interest refinement is proposed to learn the subtle nuances of how interests change and shift along the timeline, leading to a more precise representation of user interests. Our extensive experiments on multiple real-world datasets validate the effectiveness of the proposed solution, demonstrating a significant improvement over current state-of-the-art models on these benchmarks. The code is available at https://github.com/WHUIR/Horae .

Sequential recommendation systems aim to predict users’ next likely interaction based on their history. However, these systems face data sparsity and cold-start problems. Utilizing data from other domains, known as multi-domain methods, is useful for alleviating these problems. However, traditional multi-domain methods rely on meaningless ID-based item representation, which makes it difficult to align items with similar meanings from different domains, yielding sup-optimal knowledge transfer. This article introduces LLM-Rec , a framework that utilizes pre-trained Large Language Models (LLMs) for domain-agnostic recommendation. Specifically, we mix user’s behaviors from multiple domains and concatenate item titles into a sentence, then use LLMs for generating user and item representations. By mixing behaviors across different domains, we can exploit the knowledge encoded in LLMs to bridge the semantic across over multi-domain behaviors, thus obtaining semantically rich representations and improving performance in all domains. Furthermore, we explore the underlying reasons why LLMs are effective and investigate whether LLMs can understand the semantic correlations as the recommendation model, and if advanced techniques like scaling laws in NLP also work in recommendations. We conduct extensive experiments with LLMs ranging from 40 M to 6.7 B to answer the above questions and to verify the effectiveness of LLM-Rec in multi-domain recommendation. The source code is available at https://github.com/WHUIR/LLMRec .

Accurate prediction of coupon usage is crucial for promoting user consumption through targeted coupon recommendations. However, in real-world coupon recommendations, the coupon allocation process is not solely determined by the model trained with the history interaction data but is also interfered with by marketing tactics desired to fulfill specific commercial goals.This interference creates an imbalance in the interactions, which causes the data to deviate from the user's natural preferences. We refer to this deviation as the matching bias. Such biased interaction data affects the efficacy of the model, and thus it is necessary to employ debiasing techniques to prevent any negative impact. We investigate the mitigation of matching bias in coupon recommendations from a causal-effect perspective. By treating the attributes of users and coupons associated with marketing tactics as confounders, we find the confounders open the backdoor path between user-coupon matching and the conversion, which introduces spurious correlation. To remove the bad effect, we propose a novel training paradigm named Backdoor Adjustment via Group Adaptation (BAGA) for debiased coupon recommendations, which performs intervened training and inference, i.e., separately modeling each user-coupon group pair. However, modeling all possible group pairs greatly increases the computational complexity and cost. To address the efficiency challenge, we further present a simple but effective dual-tower multi-task framework and leverage the Customized Gate Control (CGC) model architecture, which separately models each user and coupon group with a separate expert module. We instantiate BAGA on five representative models: FM, DNN, NCF, MASKNET, and DEEPFM, and conduct comprehensive offline and online experiments to demonstrate the efficacy of our proposed paradigm.

The recent advancements in Deep Reinforcement Learning (DRL) have significantly enhanced the performance of adaptive Traffic Signal Control (TSC). However, DRL policies are typically represented by neural networks, which are over-parameterized black-box models. As a result, the learned policies often lack interpretability, and cannot be deployed directly in the real-world edge hardware due to resource constraints. In addition, the DRL methods often exhibit limited generalization performance, struggling to generalize the learned policy to other geographical regions. These factors limit the practical application of learning-based approaches. To address these issues, we suggest the use of an inherently interpretable program for representing the control policy. We present a new approach, Programmatic Interpretable reinforcement learning for traffic signal control (π-light), designed to autonomously discover non-differentiable programs. Specifically, we define a Domain Specific Language (DSL) and transformation rules for constructing programs, and utilize Monte Carlo Tree Search (MCTS) to find the optimal program in a discrete space. Extensive experiments demonstrate that our method consistently outperforms baseline approaches. Moreover, π-Light exhibits superior generalization capabilities compared to DRL, enabling training and evaluation across intersections from different cities. Finally, we analyze how the learned program policies can directly deploy on edge devices with extremely limited resources.

The stock market is a crucial component of the financial system, but predicting the movement of stock prices is challenging due to the dynamic and intricate relations arising from various aspects such as economic indicators, financial reports, global news, and investor sentiment. Traditional sequential methods and graph-based models have been applied in stock movement prediction, but they have limitations in capturing the multifaceted and temporal influences in stock price movements. To address these challenges, the Multi-relational Dynamic Graph Neural Network (MDGNN) framework is proposed, which utilizes a discrete dynamic graph to comprehensively capture multifaceted relations among stocks and their evolution over time. The representation generated from the graph offers a complete perspective on the interrelationships among stocks and associated entities. Additionally, the power of the Transformer structure is leveraged to encode the temporal evolution of multiplex relations, providing a dynamic and effective approach to predicting stock investment. Further, our proposed MDGNN framework achieves the best performance in public datasets compared with the state-of-the-art stock investment methods.

Recommendation systems aim to provide users with relevant suggestions, but often lack interpretability and fail to capture higher-level semantic relationships between user behaviors and profiles. In this paper, we propose a novel approach that leverages large language models (LLMs) to construct personalized reasoning graphs. These graphs link a user's profile and behavioral sequences through causal and logical inferences, representing the user's interests in an interpretable way. Our approach, LLM reasoning graphs (LLMRG), has four components: chained graph reasoning, divergent extension, self-verification and scoring, and knowledge base self-improvement. The resulting reasoning graph is encoded using graph neural networks, which serves as additional input to improve conventional recommender systems, without requiring extra user or item information. Our approach demonstrates how LLMs can enable more logical and interpretable recommender systems through personalized reasoning graphs. LLMRG allows recommendations to benefit from both engineered recommendation systems and LLM-derived reasoning graphs. We demonstrate the effectiveness of LLMRG on benchmarks and real-world scenarios in enhancing base recommendation models.

Time series forecasts of different temporal granularity are widely used in real-world applications, e.g., sales prediction in days and weeks for making different inventory plans. However, these tasks are usually solved separately without ensuring coherence, which is crucial for aligning downstream decisions. Previous works mainly focus on ensuring coherence with some straightforward methods, e.g., aggregation from the forecasts of fine granularity to the coarse ones, and allocation from the coarse granularity to the fine ones. These methods merely take the temporal hierarchical structure to maintain coherence without improving the forecasting accuracy. In this paper, we propose a novel granularity message-passing mechanism (GMP) that leverages temporal hierarchy information to improve forecasting performance and also utilizes an adaptive reconciliation (AR) strategy to maintain coherence without performance loss. Furthermore, we introduce an optimization module to achieve task-based targets while adhering to more real-world constraints. Experiments on real-world datasets demonstrate that our framework (GMP-AR) achieves superior performances on temporal hierarchical forecasting tasks compared to state-of-the-art methods. In addition, our framework has been successfully applied to a real-world task of payment traffic management in Alipay by integrating with the task-based optimization module.

Cross-Domain Sequential Recommendation (CDSR) aims to enhance personalized user experiences by leveraging user behaviors across multiple domains. Existing methods primarily focus on fusing information from various domains and modeling global user preferences, but often struggle with negative transfer, where knowledge from one domain impairs recommendation performance in another. For example, a user may enjoy watching sports games in the video domain but have no interest in participating in sports activities. Consequently, this interest does not extend to purchasing related sports gear. In such cases, a recommendation system suggesting sports gear based on the user’s viewing preferences may not elicit a positive response. To tackle this issue, we propose a novel method called Multi-Interest Bridge Recommender (MIBR). In light of the cross-domain scenario, where user preferences are not entirely consistent across domains, we design a Multi-Interest Extraction (MIE) module to capture the diversity of user interests based on a soft clustering approach. In the meantime, we design a cross-domain bridging (CDB) module, with the goal of mitigating the issue of negative transfer. CDB leverages the extracted interests as a bridge for inter-domain information transfer, enabling each domain to adaptively extract relevant information from diverse interests while ignoring unrelated ones. Extensive experiments on three popular datasets reveal MIBR’s significant superiority over baselines, e.g., with up to a 59.27% uplift in terms of HR@10 over C2DSR on the Movie-Book dataset.

Accurate workload forecasting is critical for efficient resource management in cloud computing systems, enabling effective scheduling and autoscaling. Despite recent advances with transformer-based forecasting models, challenges remain due to the non-stationary, nonlinear characteristics of workload time series and the long-term dependencies. In particular, inconsistent performance between long-term history and near-term forecasts hinders long-range predictions. This paper proposes a novel framework leveraging self-supervised multiscale representation learning to capture both long-term and near-term workload patterns. The long-term history is encoded through multiscale representations while the near-term observations are modeled via temporal flow fusion. These representations of different scales are fused using an attention mechanism and characterized with normalizing flows to handle non-Gaussian/non-linear distributions of time series. Extensive experiments on 9 benchmarks demonstrate superiority over existing methods.

Covariance matrix estimation is an important problem in statistics, with wide applications in finance, neuroscience, meteorology, oceanography, and other fields. However, when the data are high-dimensional and constantly generated and updated in a streaming fashion, the covariance matrix estimation faces huge challenges, including the curse of dimensionality and limited memory space. The existing methods either assume sparsity, ignoring any possible common factor among the variables, or obtain poor performance in recovering the covariance matrix directly from sketched data. To address these issues, we propose a novel method - KEEF: Knowledge-based Time and Memory Efficient Covariance Estimator in Factor Model and its extended variation. Our method leverages historical data to train a knowledge-based sketch matrix, which is used to accelerate the factor analysis of streaming data and directly estimates the covariance matrix from the sketched data. We provide theoretical guarantees, showing the advantages of our method in terms of time and space complexity, as well as accuracy. We conduct extensive experiments on synthetic and real-world data, comparing KEEF with several state-of-the-art methods, demonstrating the superior performance of our method.

Recently, the item textual information has been exploited with pre-trained language models (PLMs) to enrich the representations of tail items. The underlying idea is to align the hot items and tail items in terms of the external semantic knowledge covered by the PLM. However, it is non-trivial to eliminate the popularity bias by exploiting the textual semantics. One major obstacle is that the model supervision still counts on the sparse yet binary user behaviors. In the preliminary investigation, we discover that text-based recommendations also suffer from the popularity bias.

Temporal Knowledge Graph (TKG) reasoning is a crucial task that aims to predict future facts based on historical information. In the process of reasoning over TKGs, we identify two types of facts that need to be predicted: 1) recurring facts and 2) unknown facts. While existing models emphasize reasoning about recurring facts, they inadvertently overlook the importance of unknown facts. To make better predictions on both facts, we introduce a novel TKG reasoning model, named Multi-view Recurrent Network (MV-NET), which generates different views to capture reasoning patterns for both recurring and unknown facts. Specifically, MV-NET comprises three views: a recurring history view that captures repetitive features, an exploring history view that focuses on exploring new information for unknown facts, and a full history view that assimilates historical information comprehensively. Then, the historical information of each view is encoded by a multi-view recurrent network. To better integrate the embeddings of three views, we employ an adaptive scoring module, which consists of a query-aware attentive fusion mechanism to incorporate the predicted scores from three views, thus obtaining fused scores for prediction. Extensive experiments on three commonly used datasets demonstrate the superiority of MV-NET compared to many state-of-the-art baselines.

With the escalating complexity and investment cost of training deep neural networks, safeguarding them from unauthorized usage and intellectual property theft has become imperative. Especially the rampant misuse of prediction APIs to replicate models without access to the original data or architecture poses grave security threats. Diverse defense strategies have emerged to address these vulnerabilities, yet these defenses either incur heavy inference overheads or assume idealized attack scenarios. To address these challenges, we revisit the utilization of noise transition matrix as an efficient perturbation technique, which injects noise into predicted posteriors in a linear manner and integrates seamlessly into existing systems with minimal overhead, for model stealing defense. Provably, with such perturbed posteriors, the attacker's cloning process degrades into learning from noisy data. Toward optimizing the noise transition matrix, we proposed a novel bi-level optimization training framework, which performs fidelity on the victim model while the surrogate model adversarially. Comprehensive experimental results demonstrate that our method effectively thwarts model stealing attacks and achieves minimal utility tradeoffs, outperforming existing state-of-the-art defenses.

Reinforcement learning from human feedback (RLHF) and AI-generated feedback (RLAIF) have become prominent techniques that significantly enhance the functionality of pre-trained language models (LMs).These methods harness feedback, sourced either from humans or AI, as direct rewards or to shape reward models that steer LM optimization.Nonetheless, the effective integration of rewards from diverse sources presents a significant challenge due to their disparate characteristics.To address this, recent research has developed algorithms incorporating strategies such as weighting, ranking, and constraining to handle this complexity.Despite these innovations, a bias toward disproportionately high rewards can still skew the reinforcement learning process and negatively impact LM performance.This paper explores a methodology for reward composition that enables simultaneous improvements in LMs across multiple dimensions.Inspired by fairness theory, we introduce a training algorithm that aims to reduce Disparity and enhance Stability among various rewards.Our method treats the aggregate reward as a dynamic weighted sum of individual rewards, with alternating updates to the weights and model parameters.For efficient and straightforward implementation, we employ an estimation technique rooted in the mirror descent method for weight updates, eliminating the need for gradient computations.The empirical results under various types of rewards across a wide range of scenarios demonstrate the effectiveness of our method.

Neural temporal point processes(TPPs) have shown promise for modeling continuous-time event sequences. However, capturing the interactions between events is challenging yet critical for performing inference tasks like forecasting on event sequence data. Existing TPP models have focused on parameterizing the conditional distribution of future events but struggle to model event interactions. In this paper, we propose a novel approach that leverages Neural Relational Inference (NRI) to learn a relation graph that infers interactions while simultaneously learning the dynamics patterns from observational data. Our approach, the Contrastive Relational Inference-based Hawkes Process (CRIHP), reasons about event interactions under a variational inference framework. It utilizes intensity-based learning to search for prototype paths to contrast relationship constraints. Extensive experiments on three real-world datasets demonstrate the effectiveness of our model in capturing event interactions for event sequence modeling tasks.

Many distributed training techniques like Parameter Server and AllReduce have been proposed to take advantage of the increasingly large data and rich features. However, stragglers frequently occur in distributed training due to resource contention and hardware heterogeneity, which significantly hampers the training efficiency. Previous works only address part of the stragglers and could not adaptively solve various stragglers in practice. Additionally, it is challenging to use a systematic framework to address all stragglers because different stragglers require diverse data allocation and fault-tolerance mechanisms. Therefore, this paper proposes a unified distributed training framework called AntDT (Ant Distributed Training Framework) to adaptively solve the straggler problems. Firstly, the framework consists of four components, including the Stateful Dynamic Data Sharding service, Monitor, Controller, and Agent. These components work collaboratively to efficiently distribute workloads and provide a range of pre-defined straggler mitigation methods with fault tolerance, thereby hiding messy details of data allocation and fault handling. Secondly, the framework provides a high degree of flexibility, allowing for the customization of straggler mitigation solutions based on the specific circumstances of the cluster. Leveraging this flexibility, we introduce two straggler mitigation solutions, namely AntDT-ND for non-dedicated clusters and AntDT-DD for dedicated clusters, as practical examples to resolve various types of stragglers at Ant Group. Justified by our comprehensive experiments and industrial deployment statistics, AntDT outperforms other SOTA methods more than 3 × in terms of training efficiency. Additionally, in Alipay's homepage recommendation scenario, using AntDT reduces the training duration of the ranking model from 27.8 hours to just 5.4 hours.

Time series forecasting is widely used in extensive applications, such as traffic planning and weather forecasting. However, real-world time series usually present intricate temporal variations, making forecasting extremely challenging. Going beyond the mainstream paradigms of plain decomposition and multiperiodicity analysis, we analyze temporal variations in a novel view of multiscale-mixing, where time series present distinct patterns in different sampling scales. Specifically, the microscopic and the macroscopic information are reflected in fine and coarse scales, respectively, and thereby complex variations are inherently disentangled. Based on this observation, we propose TimeMixer as a fully MLP-based architecture with Past-Decomposable-Mixing (PDM) and Future-Multipredictor-Mixing (FMM) blocks to take full advantage of disentangled multiscale series in both past extraction and future prediction phases. Concretely, PDM applies the decomposition to multiscale series and further mixes the decomposed seasonal and trend components in fine-to-coarse and coarse-to-fine directions separately, which successively aggregates the microscopic seasonal and macroscopic trend information. FMM further ensembles multiple predictors to utilize complementary forecasting capabilities in multiscale observations. Consequently, our proposed TimeMixer is able to achieve consistent state-of-the-art performances in both long-term and short-term forecasting tasks with favorable run-time efficiency.

Continuous-time event sequences play a vital role in real-world domains such as healthcare, finance, online shopping, social networks, and so on. To model such data, temporal point processes (TPPs) have emerged as the most natural and competitive models, making a significant impact in both academic and application communities. Despite the emergence of many powerful models in recent years, there hasn't been a central benchmark for these models and future research endeavors. This lack of standardization impedes researchers and practitioners from comparing methods and reproducing results, potentially slowing down progress in this field. In this paper, we present EasyTPP, the first central repository of research assets (e.g., data, models, evaluation programs, documentations) in the area of event sequence modeling. Our EasyTPP makes several unique contributions to this area: a unified interface of using existing datasets and adding new datasets; a wide range of evaluation programs that are easy to use and extend as well as facilitate reproducible research; implementations of popular neural TPPs, together with a rich library of modules by composing which one could quickly build complex models. We will actively maintain this benchmark and welcome contributions from other researchers and practitioners. Our benchmark will help promote reproducible research in this field, thus accelerating research progress as well as making more significant real-world impacts. The code and data are available at \url{https://github.com/ant-research/EasyTemporalPointProcess}.

Chain-of-thought distillation is a powerful technique for transferring reasoning abilities from large language models (LLMs) to smaller student models. Previous methods typically require the student to mimic the step-by-step rationale produced by LLMs, often facing the following challenges: (i) Tokens within a rationale vary in significance, and treating them equally may fail to accurately mimic keypoint tokens, leading to reasoning errors. (ii) They usually distill knowledge by consistently predicting all the steps in a rationale, which falls short in distinguishing the learning order of step generation. This diverges from the human cognitive progression of starting with easy tasks and advancing to harder ones, resulting in sub-optimal outcomes. To this end, we propose a unified framework, called KPOD, to address these issues. Specifically, we propose a token weighting module utilizing mask learning to encourage accurate mimicry of keypoint tokens by the student during distillation. Besides, we develop an in-rationale progressive distillation strategy, starting with training the student to generate the final reasoning steps and gradually extending to cover the entire rationale. To accomplish this, a weighted token generation loss is proposed to assess step reasoning difficulty, and a value function is devised to schedule the progressive distillation by considering both step difficulty and question diversity. Extensive experiments on four reasoning benchmarks illustrate our KPOD outperforms previous methods by a large margin.

The strong performance of neural networks typically hinges on the availability of extensive labeled data, yet acquiring ground-truth labels is often challenging. Instead, noisy supervisions from multiple sources, e.g., by multiple well-designed rules, are more convenient to collect. In this paper, we focus on the realistic problem of learning from multiple noisy label sources, and argue that prior studies have overlooked the crucial self-cognition ability of neural networks, i.e., the inherent capability of autonomously distinguishing noise during training. We theoretically analyze this ability of neural networks when meeting multiple noisy label sources, which reveals that neural networks possess the capability to recognize both instance-wise noise within each single noisy label source and annotator-wise quality among multiple noisy label sources. Inspired by the theoretical analyses, we introduce an approach named Self-cognitive Denoising for Multiple noisy label sources (SDM), which exploits the self-cognition ability of neural networks to denoise during training. Furthermore, we build a selective distillation module following the theoretical insights to optimize computational efficiency. The experiments on various datasets demonstrate the superiority of our method.

Bayesian flow networks (BFNs) iteratively refine the parameters, instead of the samples in diffusion models (DMs), of distributions at various noise levels through Bayesian inference. Owing to its differentiable nature, BFNs are promising in modeling both continuous and discrete data, while simultaneously maintaining fast sampling capabilities. This paper aims to understand and enhance BFNs by connecting them with DMs through stochastic differential equations (SDEs). We identify the linear SDEs corresponding to the noise-addition processes in BFNs, demonstrate that BFN’s regression losses are aligned with denoise score matching, and validate the sampler in BFN as a first-order solver for the respective reverse-time SDE. Based on these findings and existing recipes of fast sampling in DMs, we propose specialized solvers for BFNs that markedly surpass the original BFN sampler in terms of sample quality with a limited number of function evaluations (e.g., 10) on both image and text datasets. Notably, our best sampler achieves an increase in speed of $5\sim20$ times for free.

Vision-based localization plays an important role in numerous emerging applications, including mobile robots, UAVs, AR/VR, etc. The ORB algorithm is a very classic algorithm in visual localization, typically consisting of feature point extraction and matching. Owing to the extensive computational load involved in these processes, real-time performance remains difficult in applications based on embedded devices. In this work, we present a high-performance ORB accelerator through the algorithm and hardware co-design, including a three-levels parallel processing architecture to improve performance, a pixel-level RS-BRIEF descriptor generator to reduce the computational complexity, and an on-chip sliding storage bucket matching and overwriting strategy for storage optimization. The proposed work has been implemented on the ZCU104 FPGA board achieving a high-performance (108 FPS) and a low average trajectory error at the cost of slightly fewer hardware resources compared with some SOTA works.

Customer segmentation plays a crucial role in credit risk assessment by dividing users into specific risk levels based on their credit scores. Previous methods fail to comprehensively consider the stability in the segmentation process, resulting in frequent changes and inconsistencies in users' risk levels over time. This increases potential risks to a company. To this end, this paper at first introduces and formalizes the concept of stability regret in the segmentation process. However, evaluating stability is challenging due to its black-box nature and the computational burden posed by vast user data sets. To address these challenges, this paper proposes a large-scale stable customer segmentation approach named LASCA. LASCA consists of two phases: high-quality dataset construction (HDC) and reliable data-driven optimization (RDO). Specifically, HDC utilizes an evolutionary algorithm to collect high-quality binning solutions. RDO subsequently builds a reliable surrogate model to search for the most stable binning solution based on the collected dataset. Extensive experiments conducted on real-world large-scale datasets (up to 0.8 billion) show that LASCA surpasses the state-of-the-art binning methods in finding the most stable binning solution. Notably, HDC greatly enhances data quality by 50%. RDO efficiently discovers more stable binning solutions with a 36% improvement in stability, accelerating the optimization process by 25 times via data-driven evaluation. Currently, LASCA has been successfully deployed in the large-scale credit risk assessment system of Alipay.

The fraudulent insurance claim is critical for the insurance industry.Insurance companies or agency platforms aim to confidently estimate the fraud risk of claims by gathering data from various sources.Although more data sources can improve the estimation accuracy, they inevitably lead to increased costs.Therefore, a great challenge of fraud risk verification lies in well balancing these two aspects.To this end, this paper proposes a framework named cost-efficient fraud risk optimization with submodularity (CEROS) to optimize the process of fraud risk verification.CEROS efficiently allocates investigation resources across multiple information sources, balancing the trade-off between accuracy and cost.CEROS consists of two parts that we propose: a submodular set-wise classification model * Equal Contribution.

Cross-Domain Recommendation (CDR) has been proposed to improve the recommendation accuracy in the target domain (the sparser dataset) by benefiting from the auxiliary information transferred or the knowledge learned from one or many source domains (the denser datasets). However, most of the existing CDR approaches still suffer from the problem of negative transfer caused by undifferentiated knowledge transfer, and thus the recommendation accuracy in some domains, especially in the sparser domains, is still too low, which is not practical in real application scenarios. To address this problem, we propose a novel Active Masked Attention framework, i.e., AMA-CDR, for many-to-many CDR scenarios. Our AMA-CDR pursues a higher goal for CDR approaches, i.e., improving the recommendation performance in the target domain to achieve a practically usable level, which is meaningful and challenging in real CDR systems. Specifically, AMA-CDR adopts an end-to-end graph embedding to reduce the objective distortion between graph embedding and embedding combination. More importantly, we propose an active mask for the embedding combination to ease negative transfer, which leverages both the prior knowledge, i.e., data density, and the posterior knowledge, i.e., sample uncertainty. Extensive experiments conducted on two public datasets demonstrate that our proposed AMA-CDR models significantly outperform the state-of-the-art approaches and achieve the new goal.

This paper considers the problem of learning from multiple sets of inaccurate labels, which can be easily obtained from low-cost annotators, such as rule-based annotators. Previous works typically concentrate on aggregating information from all the annotators, overlooking the significance of data refinement. This paper presents a collaborative refining approach for learning from inaccurate labels. To refine the data, we introduce the annotator agreement as an instrument, which refers to whether multiple annotators agree or disagree on the labels for a given sample. For samples where some annotators disagree, a comparative strategy is proposed to filter noise. Through theoretical analysis, the connections among multiple sets of labels, the respective models trained on them, and the true labels are uncovered to identify relatively reliable labels. For samples where all annotators agree, an aggregating strategy is designed to mitigate potential noise. Guided by theoretical bounds on loss values, a sample selection criterion is introduced and modified to be more robust against potentially problematic values. Through these two methods, all the samples are refined during training, and these refined samples are used to train a lightweight model simultaneously. Extensive experiments are conducted on benchmark and real-world datasets to demonstrate the superiority of our methods.

Gradient-based bilevel programming leverages unrolling differentiation (UD) or implicit function theorem (IFT) to solve hyperparameter optimization (HO) problems, and is proven effective and scalable in practice. To understand their generalization behavior, existing works establish upper bounds on the uniform stability of these algorithms, while their tightness is still unclear. To this end, this paper attempts to establish stability lower bounds for UD-based and IFT-based algorithms. A central technical challenge arises from the dependency of each outer-level update on the concurrent stage of inner optimization in bilevel programming. To address this problem, we introduce lower-bounded expansion properties to characterize the instability in update rules which can serve as general tools for lower-bound analysis. These properties guarantee the hyperparameter divergence at the outer level and the Lipschitz constant of inner output at the inner level in the context of HO. Guided by these insights, we construct a quadratic example that yields tight lower bounds for the UD-based algorithm and meaningful bounds for a representative IFT-based algorithm. Our tight result indicates that uniform stability has reached its limit in stability analysis for the UD-based algorithm.

Recently, various strategies for distributed training of large language models (LLMs) have been proposed. By categorizing them into basic strategies and composite strategies, we have discovered that existing basic strategies provide limited options in specific scenarios, leaving considerable room for optimization in training speed. In this paper, we rethink the impact of memory and communication costs on the training speed of LLMs, taking into account the impact of intra- and inter-group communication performance disparities, and then propose a new set of basic strategies named the \textbf{Pa}rtial \textbf{R}edundancy \textbf{O}ptimizer (PaRO). PaRO Data Parallelism (PaRO-DP) accelerates LLM training through refined model state partitioning and tailored training procedures. At the same time, PaRO Collective Communications (PaRO-CC) speeds up collective communication operations by rearranging the topology. We also propose a guideline for choosing different DP strategies based on simple quantitative calculations, which yields minimal ranking errors. Our experiments demonstrate that PaRO improves the training speed of LLMs by up to 266\% that of ZeRO-3 as basic DP strategies. Moreover, employing PaRO-CC independently for model parallel strategies, such as Megatron, can also boost the training speed by 17\%.

Recently, Large Language Models (LLMs) have shown significant potential in addressing the isolation issues faced by recommender systems. However, despite performance comparable to traditional recommenders, the current methods are cost-prohibitive for industrial applications. Consequently, existing LLM-based methods still need to catch up regarding effectiveness and efficiency. To tackle the above challenges, we present an LLM-enhanced recommendation framework named ReLand, which leverages Retrieval to effortlessly integrate Large language models’ insights into industrial recommenders. Specifically, ReLand employs LLMs to perform generative recommendations on sampled users (a.k.a., seed users), thereby constructing an LLM Reasoning Pool. Subsequently, we leverage retrieval to attach reliable recommendation rationales for the entire user base, ultimately effectively improving recommendation performance. Extensive offline and online experiments validate the effectiveness of ReLand. Since January 2024, ReLand has been deployed in the recommender system of Alipay, achieving statistically significant improvements of 3.19% in CTR and 1.08% in CVR.

With the rise of large language models (LLMs), recent works have leveraged LLMs to improve the performance of click-through rate (CTR) prediction. However, we argue that a critical obstacle remains in deploying LLMs for practical use: the efficiency of LLMs when processing long textual user behaviors. As user sequences grow longer, the current efficiency of LLMs is inadequate for training on billions of users and items. To break through the efficiency barrier of LLMs, we propose Behavior Aggregated Hierarchical Encoding (BAHE) to enhance the efficiency of LLM-based CTR modeling. Specifically, BAHE proposes a novel hierarchical architecture that decouples the encoding of user behaviors from inter-behavior interactions. Firstly, to prevent computational redundancy from repeated encoding of identical user behaviors, BAHE employs the LLM's pre-trained shallow layers to extract embeddings of the most granular, atomic user behaviors from extensive user sequences and stores them in the offline database. Subsequently, the deeper, trainable layers of the LLM facilitate intricate inter-behavior interactions, thereby generating comprehensive user embeddings. This separation allows the learning of high-level user representations to be independent of low-level behavior encoding, significantly reducing computational complexity. Finally, these refined user embeddings, in conjunction with correspondingly processed item embeddings, are incorporated into the CTR model to compute the CTR scores. Extensive experimental results show that BAHE reduces training time and memory by five times for CTR models using LLMs, especially with longer user sequences. BAHE has been deployed in a real-world system, allowing for daily updates of 50 million CTR data on 8 A100 GPUs, making LLMs practical for industrial CTR prediction.

Click-through rate (CTR) prediction plays a crucial role in recommendation systems, with significant impact on user experience and platform revenue generation. Despite the various public CTR datasets available due to increasing interest from both academia and industry, these datasets have limitations. They cover a limited range of scenarios and predominantly focus on ID-based features, neglecting the vital role of multi-modal features for effective multi-scenario CTR prediction. Moreover, their scale is modest compared to real-world industrial datasets, hindering robust and comprehensive evaluation of complex models. To address these challenges, we introduce a large-scale Multi-Scenario Multi-Modal CTR dataset named AntM2 C, built from real industrial data from Alipay. This dataset offers an impressive breadth and depth of information, covering CTR data from four diverse business scenarios, including advertisements, consumer coupons, mini-programs, and videos. Unlike existing datasets, AntM2 C provides not only ID-based features but also five textual features and one image feature for both users and items, supporting more delicate multi-modal CTR prediction. AntM2 C is also substantially larger than existing datasets, comprising 100 million CTR data. This scale allows for robust and comprehensive evaluation and comparison of CTR prediction models. We employ AntM2 C to construct several typical CTR tasks, including multi-scenario modeling, item and user cold-start modeling, and multi-modal modeling. Initial experiments and comparisons with baseline methods have shown that AntM2 C presents both new challenges and opportunities for CTR models, with the potential to significantly advance CTR research. The AntM2 C dataset is available at https://www.atecup.cn/OfficalDataSet.

Recommender systems are widely used in various online platforms. In the context of sequential recommendation, it is essential to accurately capture the chronological patterns in user activities to generate relevant recommendations. Conventional ID-based sequential recommenders have shown promise but lack comprehensive real-world knowledge about items, limiting their effectiveness. Recent advancements in Large Language Models (LLMs) offer the potential to bridge this gap by leveraging the extensive real-world knowledge encapsulated in LLMs. However, integrating LLMs into sequential recommender systems comes with its own challenges, including inadequate representation of sequential behavior patterns and long inference latency. In this paper, we propose SeRALM (Enhancing Sequential Recommenders with Augmented Knowledge from Aligned Large Language Models) to address these challenges. SeRALM integrates LLMs with conventional ID-based sequential recommenders for sequential recommendation tasks. We combine text-format knowledge generated by LLMs with item IDs and feed this enriched data into ID-based recommenders, benefitting from the strengths of both paradigms. Moreover, we develop a theoretically underpinned alignment training method to refine LLMs' generation using feedback from ID-based recommenders for better knowledge augmentation. We also present an asynchronous technique to expedite the alignment training process. Experimental results on public benchmarks demonstrate that SeRALM significantly improves the performances of ID-based sequential recommenders. Further, a series of ablation studies and analyses corroborate SeRALM's proficiency in steering LLMs to generate more pertinent and advantageous knowledge across diverse scenarios.

The importance of incorporating interpretability into machine learning models has been increasingly emphasized. While previous literature has typically focused on feature level interpretability, such as analyzing which features are important and how they influence the final decision, real-world applications often require domain level interpretability, which relates to a group of features. Domain-level interpretability holds the potential for enhanced informativeness and comprehensibility. Unfortunately, there has been limited research in this direction. In this paper, we address this issue and introduce our proposed method DIDE, which obtains domain-level interpretability from domain-specific latent embeddings. To enhance the effectiveness of the framework, we draw inspiration from the gradient smooth philosophy and propose noisy injection in the embedding space, resulting in smoothed interpretability. We conduct extensive experiments to validate the effectiveness of DIDE, and demonstrate its applications in assisting daily business tasks in Alipay.

To help merchants/customers to provide/access a variety of services through miniapps, online service platforms have occupied a critical position in the effective content delivery, in which how to recommend items in the new domain launched by the service provider for customers has become more urgent. However, the non-negligible gap between the source and diversified target domains poses a considerable challenge to cross-domain recommendation systems, which often leads to performance bottlenecks in industrial settings. While entity graphs have the potential to serve as a bridge between domains, rudimentary utilization still fail to distill useful knowledge and even induce the negative transfer issue. To this end, we propose PEACE, a Prototype lEarning Augmented transferable framework for Cross-domain rEcommendation. For domain gap bridging, PEACE is built upon a multi-interest and entity-oriented pre-training architecture which could not only benefit the learning of generalized knowledge in a multi-granularity manner, but also help leverage more structural information in the entity graph. Then, we bring the prototype learning into the pre-training over source domains, so that representations of users and items are greatly improved by the contrastive prototype learning module and the prototype enhanced attention mechanism for adaptive knowledge utilization. To ease the pressure of online serving, PEACE is deployed in a lightweight manner, and significant performance improvements are observed in both online and offline environments.

The rapid development of graph neural networks (GNNs) encourages the rising of link prediction, achieving promising performance with various applications. Unfortunately, through a comprehensive analysis, we surprisingly find that current link predictors with dynamic negative samplers (DNSs) suffer from the migration phenomenon between ''easy'' and ''hard'' samples, which goes against the preference of DNS of choosing "hard" negatives, thus severely hindering capability. Towards this end, we propose the MeBNS framework, serving as a general plugin that can potentially improve current negative sampling based link predictors. In particular, we elaborately devise a Meta-learning Supported Teacher-student GNN (MST-GNN) that is not only built upon teacher-student architecture for alleviating the migration between ''easy'' and ''hard'' samples but also equipped with a meta learning based sample re-weighting module for helping the student GNN distinguish ''hard'' samples in a fine-grained manner. To effectively guide the learning of MST-GNN, we prepare a Structure enhanced Training Data Generator (STD-Generator) and an Uncertainty based Meta Data Collector (UMD-Collector) for supporting the teacher and student GNN, respectively. Extensive experiments show that the MeBNS achieves remarkable performance across six link prediction benchmark datasets.

Causal effect estimation from networked observational data encounters notable challenges, primarily hidden confounders arising from network structure, or spillover effects that influence unit's outcomes based on neighboring treatment assignments. Existing graph neural network (GNN)-based methods have endeavored to address these challenges, utilizing the GNN's message-passing mechanism to capture hidden confounders or model spillover effects. However, they mainly focus on transductive causal effect learning on a single networked data, limiting their efficacy in inductive settings for real-world applications where networked data often originates from multiple environments influenced by potentially varying time or geographical regions. In light of this, we introduce the principle of invariance to the task of causal effect estimation on networked data, culminating in our Invariant Graph Learning (IGL) framework. Specifically, it first generates multiple networked data to simulate diverse environments from a given observational data. Then it further encourages the model to learn environment-invariant representations for confounders and spillover effects. Such a design enables the model to extrapolate beyond a single observed environment, thereby improving the performance of causal effect estimation in potential new environments. Extensive experiments on two real-world datasets demonstrates the superiority of our approach.

Large language models (LLMs) open up new horizons for sequential recommendations, owing to their remarkable language comprehension and generation capabilities. However, there are still numerous challenges that should be addressed to successfully implement sequential recommendations empowered by LLMs. Firstly, user behavior patterns are often complex, and relying solely on one-step reasoning from LLMs may lead to incorrect or task-irrelevant responses. Secondly, the prohibitively resource requirements of LLM (e.g., ChatGPT-175B) are overwhelmingly high and impractical for real sequential recommender systems. In this paper, we propose a novel Step-by-step knowLedge dIstillation fraMework for recommendation (SLIM), paving a promising path for sequential recommenders to enjoy the exceptional reasoning capabilities of LLMs in a "slim" (i.e. resource-efficient) manner. We introduce CoT prompting based on user behavior sequences for the larger teacher model. The rationales generated by the teacher model are then utilized as labels to distill the downstream smaller student model (e.g., LLaMA2-7B). In this way, the student model acquires the step-by-step reasoning capabilities in recommendation tasks. We encode the generated rationales from the student model into a dense vector, which empowers recommendation in both ID-based and ID-agnostic scenarios. Extensive experiments demonstrate the effectiveness of SLIM over state-of-the-art baselines, and further analysis showcasing its ability to generate meaningful recommendation reasoning at affordable costs.

Graph neural networks (GNNs) have gained popularity in modeling various complex networks, e.g., social network and webpage network. Despite the promising accuracy, the confidences of GNNs are shown to be miscalibrated, indicating limited awareness of prediction uncertainty and harming the reliability of model decisions. Existing calibration methods primarily focus on improving GNN models, e.g., adding regularization during training or introducing temperature scaling after training. In this paper, we argue that the miscalibration of GNNs may stem from the graph data and can be alleviated through topology modification. To support this motivation, we conduct data observations by examining the impacts ofdecisive andhomophilic edges on calibration performance, where decisive edges play a critical role in GNN predictions and homophilic edges connect nodes of the same class. By assigning larger weights to these edges in the adjacency matrix, we observe an improvement in calibration performance without sacrificing classification accuracy. This suggests the potential of a data-centric approach for calibrating GNNs. Motivated by our observations, we propose Data-centric Graph Calibration (DCGC), which uses two edge weighting modules to adjust the input graph for GNN calibration. The first module learns the weights of decisive edges by parameterizing the adjacency matrix and enabling backpropagation of the prediction loss to edge weights. This emphasizes critical edges that fit the prediction needs. The second module computes weights for homophilic edges based on predicted label distributions, assigning larger weights to edges with stronger homophily. These modifications operate at the data level and can be easily integrated with temperature scaling-based methods for better calibration. Experimental results on 8 benchmark datasets demonstrate that DCGC achieves state-of-the-art calibration performance, with an average relative improvement of 36.4% in ECE, while maintaining or even slightly improving classification accuracy. Ablation studies and hyper-parameter analysis further validate the effectiveness and robustness of our proposed method DCGC. Code and data are available at https://github.com/BUPT-GAMMA/DCGC.

Machine learning techniques have been widely used and are mostly performed by predicting point estimation. Nevertheless, there are various scenarios that require more information beyond only point estimation. Probabilistic prediction is a typical research topic that provides probability distribution while predicting, which addressed with much attention recently. Most of the previous works make an assumption on the probabilistic distribution to a certain extent, which may lead to potential errors in the subsequent decision-making procedure. In this paper, we propose a distribution-free method for regression problems on real-value response under the probabilistic prediction framework and present an effective boosting-based method to perform the training process. Moreover, we introduce a further improved method to accomplish an unbiased mean estimation of the target distribution. Thorough experiments on multiple benchmark data are conducted to demonstrate the effectiveness of the proposed method with regard to different measures.

Based on the diversified application scenarios at Ant Group, we built the Ant Knowledge Graph Platform (AKGP). It has constructed numerous domain-specific knowledge graphs related to merchants, companies, accounts, products, and more. AKGP manages trillions of structured knowledge graphs, serving search, recommendation, risk control and other businesses. However, as the demand increasing for various workloads such as graph pattern matching, graph representation learning, and cross-domain knowledge reuse, the existing warehouse systems based on relational DBMS or graph databases are unable to meet the requirements. To address these issues, we propose KGFabric, an industrial-scale knowledge graph management system built on the distributed file system (DFS). KGFabric offers a nearline knowledge storage engine that utilizes a Semantic-enhanced Programmable Graph (SPG) model, which is compatible with the Labeled Property Graph (LPG) model. The data is persistently stored in DFS, such as HDFS, which leverages the POSIX file system API, making it suitable for deployment in multi-cloud environment at low cost. KGFabric provides a native graph-based and hybrid storage format that can serve as a shared backend for parallel graph computing systems, significantly accelerating the analysis of multi-workload. Additionally, KGFabric includes a graph fabric framework that minimizes data duplication and guarantees data security. KGFabric is able to manage Peta-scale data and has supported graph fabric and analysis with over 100 billion relations at Ant Group. We conduct experiments on various datasets to evaluate the performance of KGFabric. Compared with popular relational DBMS and graph databases, the storage space for semantic relations is reduced by over 90%. The performance of graph fabric improves by 21× in real-world workloads. In multi-hop semantic graph analysis, KGFabric enhances performance by 100×.

Autoscaling is a critical mechanism in cloud computing, enabling the autonomous adjustment of computing resources in response to dynamic workloads. This is particularly valuable for co-located, long-running applications with diverse workload patterns. The primary objective of autoscaling is to regulate resource utilization at a desired level, effectively balancing the need for resource optimization with the fulfillment of Service Level Objectives (SLOs). Many existing proactive autoscaling frameworks may encounter prediction deviations arising from the frequent fluctuations of cloud workloads. Reactive frameworks, on the other hand, rely on realtime system feedback, but their hysteretic nature could lead to violations of stringent SLOs. Hybrid frameworks, while prevalent, often feature independently functioning proactive and reactive modules, potentially leading to incompatibility and undermining the overall decision-making efficacy. In addressing these challenges, we propose OptScaler, a collaborative autoscaling framework that integrates proactive and reactive modules through an optimization module. The proactive module delivers reliable future workload predictions to the optimization module, while the reactive module offers a self-tuning estimator for real-time updates. By embedding a Model Predictive Control (MPC) mechanism and chance constraints into the optimization module, we further enhance its robustness. Numerical results have demonstrated the superiority of our workload prediction model and the collaborative framework, leading to over a 36% reduction in SLO violations compared to prevalent reactive, proactive, or hybrid autoscalers. Notably, OptScaler has been successfully deployed at Alipay, providing autoscaling support for the world-leading payment platform.

Federated Learning (FL) is nowadays one of the most promising paradigms for privacy-preserving distributed learning. Without revealing its local private data to outsiders, a client in FL systems collaborates to build a global Deep Neural Network (DNN) by submitting its local model parameter update to a central server for iterative aggregation. With secure multi-party computation protocols, the submitted update ofany clientis also by design invisible to the server. Seemingly, this standard design is a win-win for client privacy and service provider utility. Ironically, any attacker may also use manipulated or impersonated client to submit almost any attack payloads under the umbrella of the FL protocol itself. In this work, we craft a practical backdoor attack on FL systems that is proved to be simultaneously effective and stealthy on diverse use cases of FL systems and leading commercial FL platforms in the real world. Basically, we first identify a small number of redundant neurons which tend to be rarely or slightly updated in the model, and then inject backdoor into these redundant neurons instead of the whole model. In this way, our backdoor attack can achieve a high attack success rate with a minor impact on the accuracy of the original task. As countermeasures, we further consider several common technical choices including robust aggregation mechanisms, differential privacy mechanism,s and network pruning. However, none of the defenses show desirable defense capability against our backdoor attack. Our results strongly highlight the vulnerability of existing FL systems against backdoor attacks and the urgent need to develop more effective defense mechanisms.

Recent studies have revealed that GNNs are vulnerable to adversarial attacks. Most existing robust graph learning methods measure model robustness based on label information, rendering them infeasible when label information is not available. A straightforward direction is to employ the widely used Infomax technique from typical Unsupervised Graph Representation Learning (UGRL) to learn robust unsupervised representations. Nonetheless, directly transplanting the Infomax technique from typical UGRL to robust UGRL may involve a biased assumption. In light of the limitation of Infomax, we propose a novel unbiased robust UGRL method calledRobust Graph Information Bottleneck(RGIB), which is grounded in the Information Bottleneck (IB) principle. Our RGIB attempts to learn robust node representations against adversarial perturbations by preserving the original information in the benign graph while eliminating the adversarial information in the adversarial graph. There are mainly two challenges to optimizing RGIB: 1) high complexity of adversarial attack to perturb node features and graph structure jointly in the training procedure; 2) mutual information estimation upon adversarially attacked graphs. To tackle these problems, we further propose an efficient adversarial training strategy with only feature perturbations and an effective mutual information estimator with the subgraph-level summary. Moreover, we theoretically establish a connection between our proposed RGIB and the robustness of downstream classifiers, revealing that RGIB can provide a lower bound on the adversarial risk of downstream classifiers. Extensive experiments over several benchmarks and downstream tasks demonstrate the effectiveness and superiority of our proposed method.

Sequential data mining is ubiquitous in various scenarios. Modeling event sequence and predicting event occurrence is of vital importance in sequential data mining, and Temporal Point Processes (TPP) are widely used in this area. Conventional TPP use objective functions as sum of classification loss for event type and regression loss for occurrence time, leading to practical limitations that conventional TPP is unable to predict the occurrence of each type of event and distinguish the dependency within and between different event types. To tackle these defects, we propose a Multi-task Decouple Learning (MTDL) framework to model TPP from a novel perspective of Multi-task Learning (MTL), i.e., predicting the next-step occurrence time for all event types using a weighted multi-task regression loss. We experiment with three state-of-the-arts, showing that the proposed MTDL framework can improve the performance of original TPP models. Moreover, we develop a Hierarchical Attentive Point Process (HAPP) to further exploit the potential of the proposed MTDL framework, using a hierarchical attention mechanism to capture the inner-sequence time dependency within the same type of events and the inter-sequence dependency between different types of events. Experiments on real-world business dataset and public datasets show the efficacy of the proposed method.

Social recommendation provides an auxiliary social network structure to enhance recommendation performances. By formulating user-user social network and user-item interaction graph, modern social recommendation architecture is built on learning user and item embeddings into Euclidean space with graph convolution operations. However, the Euclidean space suffers structure distortion when representing the nature power-law distribution of graphs, leading to sub-optimal results for graph based social recommendation. Recently, some studies have explored the alternative of graph embedding learning into hyperbolic space, which can preserve the hierarchy of real-world graphs. However, directly applying current hyperbolic graph embedding models for social recommendation is non-trivial as two challenges: network heterogeneity and social diffusion noise. First, due to the semantic gap existing between social networks and user-item interactions, how to tackle the heterogeneity issue of social recommendation under hyperbolic formulation? Second, explicit modeling of social diffusion easily introduces noise for user preference learning, especially for those active users with amounts of interactions. To tackle the above challenges, in this paper, we propose aHyperbolic Graph Learning based Social Recommendation (HGSR)model. Firstly, we exploit social structure with hyperbolic social embedding pre-training, which could preserve the hierarchical properties of social networks. Secondly, we construct the heterogeneous graph based on user-item interactions and social networks, then treat the pre-trained social embeddings as an additional feature input for user preference learning. Such that, we combine explicit heterogeneous graph learning and implicit feature enhancement for the hyperbolic social recommendation, which can well tackle heterogeneity and social noise issues. We conduct empirical studies on four datasets, and extensive experiments demonstrate the effectiveness of our proposed model compared to state-of-the-art baselines.

Text classification is one of the fundamental tasks in natural language processing, which requires an agent to determine the most appropriate category for input sentences. Recently, deep neural networks have achieved impressive performance in this area, especially pretrained language models (PLMs). Usually, these methods concentrate on input sentences and corresponding semantic embedding generation. However, for another essential component: labels, most existing works either treat them as meaningless one-hot vectors or use vanilla embedding methods to learn label representations along with model training, underestimating the semantic information and guidance that these labels reveal. To alleviate this problem and better exploit label information, in this article, we employ self-supervised learning (SSL) in model learning process and design a novel self-supervised relation of relation ( [Formula: see text]) classification task for label utilization from a one-hot manner perspective. Then, we propose a novel relation of relation learning network( [Formula: see text]-Net) for text classification, in which text classification and [Formula: see text] classification are treated as optimization targets. Meanwhile, triplet loss is employed to enhance the analysis of differences and connections among labels. Moreover, considering that one-hot usage is still short of exploiting label information, we incorporate external knowledge from WordNet to obtain multiaspect descriptions for label semantic learning and extend [Formula: see text]-Net to a novel description-enhanced label embedding network(DELE) from a label embedding perspective. One step further, since these fine-grained descriptions may introduce unexpected noise, we develop a mutual interaction module to select appropriate parts from input sentences and labels simultaneously based on contrastive learning (CL) for noise mitigation. Extensive experiments on different text classification tasks reveal that [Formula: see text]-Net can effectively improve the classification performance and DELE can make better use of label information and further improve the performance. As a byproduct, we have released the codes to facilitate other research.

Though neural networks have achieved impressive prediction performance, it's still hard for people to understand what neural networks have learned from the data. The black-box property of neural networks already becomes one of the main obstacles preventing from being applied to many high-stakes applications, such as finance and medicine that have critical requirement on the model transparency and interpretability. In order to enhance the explainability of neural networks, we propose a neural rule learning method-Feature Interactive Neural Rule Learning (FINRule) to incorporate the expressivity of neural networks and the interpretability of rule-based systems. Specifically, we conduct rule learning as differential discrete combination encoded by a feedforward neural network, in which each layer acts as a logical operator of explainable decision conditions. The first hidden layer can act as sharable atomic conditions which are connected to next hidden layer for formulating decision rules. Moreover, we propose to represent both atomic condition and rules with contextual embeddings, with aim to enrich the expressivity power by capturing high-order feature interactions. We conduct comprehensive experiments on real-world datasets to validate both effectiveness and explainability of the proposed method.

User behavior representation learned by self-supervised pre-training tasks is widely used in various domains and applications. Conventional methods usually follow the methodology in Natural Language Processing (NLP) to set the pre-training tasks. They either randomly mask some of the behaviors in the sequence and predict the masked ones or predict the next k behaviors. These methods fit for text sequence, in which the tokens are sequentially arranged subject to linguistic criterion. However, the user behavior sequences can be stochastic with noise and randomness. The same paradigm is intractable for learning a robust user behavioral representation.

Predictive Autoscaling is used to forecast the workloads of servers and prepare the resources in advance to ensure service level objectives (SLOs) in dynamic cloud environments. However, in practice, its prediction task often suffers from performance degradation under abnormal traffics caused by external events (such as sales promotional activities and applications' re-configurations), for which a common solution is to re-train the model with data of a long historical period, but at the expense of high computational and storage costs. To better address this problem, we propose a replay-based continual learning method, i.e., Density-based Memory Selection and Hint-based Network Learning Model (DMSHM), using only a small part of the historical log to achieve accurate predictions. First, we discover the phenomenon of sample overlap when applying replay-based continual learning in prediction tasks. In order to surmount this challenge and effectively integrate new sample distribution, we propose a density-based sample selection strategy that utilizes kernel density estimation to calculate sample density as a reference to compute sample weight, and employs weight sampling to construct a new memory set. Then we implement hint-based network learning based on hint representation to optimize the parameters. Finally, we conduct experiments on public and industrial datasets to demonstrate that our proposed method outperforms state-of-the-art continual learning methods in terms of memory capacity and prediction accuracy. Furthermore, we demonstrate remarkable practicability of DMSHM in real industrial applications.

In real-world industrial advertising systems, user behavior sparsity is a key issue that affects online recommendation performance. We observe that users with rich behaviors can obtain better recommendation results than those with sparse behaviors in a conversion-rate (CVR) prediction model. Inspired by this phenomenon, we propose a new method SeqGen, in an effort to exploit user side information to bridge the gap between rich and sparse behaviors. SeqGen is a learnable and pluggable module, which can be easily integrated into any CVR model and no longer requires two-stage training as in previous works. In particular, SeqGen learns a mapping relationship between the user side information and behavior sequences, only on the basis of the users with long behavior sequences. After that, SeqGen can generate rich sequence features for users with sparse behaviors based on their side information, so as to alleviate the issue of user behavior sparsity. The generated sequence features will then be fed into the classifier tower of an arbitrary CVR model together with the original sequence features. To the best of our knowledge, our approach constitutes the first attempt to exploit user side information for addressing the user behavior sparsity issue. We validate the effectiveness of SeqGen on the publicly available dataset MovieLens-1M, and our method receives an improvement of up to 0.5% in terms of the AUC score. More importantly, we successfully deploy SeqGen in the commercial advertising system Xlight of Alipay, which improves the grouped AUC of the CVR model by 0.6% and brings a boost of 0.49% in terms of the conversion rate on A/B testing.

Treatment effect estimation is essential in the causal inference literature, which has attracted increasing attention in recent years. Most previous methods assume that the training and test data are drawn from the same distribution, which may not hold in practice since the effect estimators may need to be deployed across domains. Meanwhile, in real-world applications, little or no targeted treatments may be conducted in the new domain. Therefore, we focus on a more realistic scenario in this paper, where treatments and outcomes can be observed in the source domain, but the target domain only contains some unlabeled data, i.e., only features are available. In this scenario, thedistribution shift exists not only in the source data due to the selection bias between the control and treated groups, but also between the source and target data. We propose a novel direct learning framework along with the distribution adaptation and reliable scoring modules. In the distribution adaptation module, we design three specialized density ratio estimators to aid the issue of complex distribution shifts. Even so, we may face the challenge of unreliable pseudo-effects in this framework. To address that, we also design the uncertainty-based reliable scoring module as a vital support, which makes the method more reliable. The experiments are conducted on synthetic data and benchmark datasets, which demonstrate the superiority of our method.

Nowadays, contrastive self-supervised learning has been widely incorporated into sequential recommender systems. However, most existing contrastive sequential recommender systems simply emphasize the overall information of interaction sequences, thereby neglecting the special periodic patterns of user behavior. In this study, we propose that users exhibit emanative periodicity towards a group of correlated items, i.e., user behavior follow a certain periodic pattern while their interests may shift from one item to other related items over time. In light of this observation, we present a hierarchical contrastive learning framework to model EmAnative periodicity for SEquential Recommendation (referred to as EASE). Specifically, we design dual-channel contrastive strategy from the perspective of correlation and periodicity to capture emanative periodic patterns. Furthermore, we extend the traditional binary contrastive loss with hierarchical constraint to handle hierarchical contrastive samples, thus preserving the inherent hierarchical information of correlation and periodicity. Comprehensive experiments conducted on five datasets substantiate the effectiveness of our proposed EASE in improving sequential recommendation.

At present, people are gradually becoming aware of financial management and thus fund recommendation attracts more and more attention to help them find suitable funds quickly. As a user usually takes many factors (e.g., fund theme, fund manager) into account when investing a fund and the fund usually consists of a substantial collection of investments, effectively modeling multi-interest representations is more crucial for personalized fund recommendation than the traditional goods recommendation. However, existing multi-interest methods are largely sub-optimal for fund recommendation, since they ignore financial domain knowledge and diverse fund investment intentions. In this work, we propose a Disentangled Interest importance aware Knowledge Graph Neural Network (DIKGNN) for personalized fund recommendation on FinTech platforms. In particular, we restrict the multiple intent spaces by introducing the attribute nodes from the fund knowledge graph as the minimum intent modeling unit to utilize financial domain knowledge and provide interpretability. In the intent space, we define disentangled intent representations, equipped with intent importance distributions to describe the diverse fund investment intentions. Then we design a new neighbor aggregation mechanism with the learned intent importance distribution upon the interaction graph and knowledge graph to collect multi-intent information. Furthermore, we leverage micro independence and macro balance constraints on the representations and distributions respectively to encourage intent independence and diversity. The extensive experiments on public recommendation benchmarks demonstrate that DIKGNN can achieve substantial improvement over state-of-the-art methods. Our proposed model is also evaluated over one real-world industrial fund dataset from a FinTech platform and has been deployed online.

Recently, a new paradigm, meta learning, has been widely applied to Deep Learning Recommendation Models (DLRM) and significantly improves statistical performance, especially in cold-start scenarios. However, the existing systems are not tailored for meta learning based DLRM models and have critical problems regarding efficiency in distributed training in the GPU cluster. It is because the conventional deep learning pipeline is not optimized for two task-specific datasets and two update loops in meta learning. This paper provides a high-performance framework for large-scale training for Optimization-based Meta DLRM models over the G PU cluster, namely G -Meta. Firstly, G-Meta utilizes both data parallelism and model parallelism with careful orchestration regarding computation and communication efficiency, to enable high-speed distributed training. Secondly, it proposes a Meta-IO pipeline for efficient data ingestion to alleviate the I/O bottleneck. Various experimental results show that G-Meta achieves notable training speed without loss of statistical performance. Since early 2022, G-Meta has been deployed in Alipay's core advertising and recommender system, shrinking the continuous delivery of models by four times. It also obtains 6.48% improvement in Conversion Rate (CVR) and 1.06% increase in CPM (Cost Per Mille) in Alipay's homepage display advertising, with the benefit of larger training samples and tasks.

Recently, the growth of service platforms brings great convenience to both users and merchants, where the service search engine plays a vital role in improving the user experience by quickly obtaining desirable results via textual queries. Unfortunately, users’ uncontrollable search customs usually bring vast amounts of long-tail queries, which severely threaten the capability of search models. Inspired by recently emerging graph neural networks (GNNs) and contrastive learning (CL), several efforts have been made in alleviating the long-tail issue and achieve considerable performance. Nevertheless, they still face a few major weaknesses. Most importantly, they do not explicitly utilize the contextual structure between heads and tails for effective knowledge transfer, and intention-level information is commonly ignored for more generalized representations.To this end, we develop a novel framework GARCIA, which exploits the graph based knowledge transfer and intention based representation generalization in a contrastive setting. In particular, we employ an adaptive encoder to produce informative representations for queries and services, as well as hierarchical structure aware representations of intentions. To fully understand tail queries and services, we equip GARCIA with a novel multi-granularity contrastive learning module, which powers representations through knowledge transfer, structure enhancement and intention generalization. Subsequently, the complete GARCIA is well trained in a pre-training&fine-tuning manner. At last, we conduct extensive experiments on both offline and online environments, which demonstrates the superior capability of GARCIA in improving tail queries and overall performance in service search scenarios.

Decision rules have been widely applied in industrial applications such as finance, medicine, and biology, due to the critical requirement of interpretability. In order to make decision rules easier and more widely used in financial scenarios, an automatic intelligent rule system with rule learning and rule management capabilities is needed. However, the rule system for financial applications has distinctive challenges both in algorithms and systems. From the algorithm perspective, due to the characteristics of the financial data and scenarios, the rule learning algorithm faces the class-imbalanced issue, the scalability issue, and the diversity of optimization objectives. From the system perspective, a flexible rule learning and management framework is needed to adapt to fast-changing financial applications with heterogenous data, and engineering optimization is required to ensure the time and space efficiency of rule learning. In this work, we focus on developing a Rule-based Decision System (RDS) to deal with the algorithmic and systematic challenges mentioned above. RDS covers the full life cycle of the decision rules, including the rule learning module, rule management module, and rule deployment module. Moreover, the rule system offers an interactive interface to allow users to integrate the expert experiences into the decision rules and realize the human-in-the-loop. The RDS has been deployed on one of the world’s largest trading and money transfer platforms, serving hundreds of millions of users and transactions.

With the rapid development of Graph Neural Networks (GNNs), more and more studies focus on system design to improve training efficiency while ignoring the efficiency of GNN inference. Actually, GNN inference is a non-trivial task, especially in industrial scenarios with giant graphs, given three main challenges, i.e., scalability tailored for full-graph inference on huge graphs, inconsistency caused by stochastic acceleration strategies (e.g., sampling), and the serious redundant computation issue. To address the above challenges, we propose a scalable system named InferTurbo to boost the GNN inference tasks in industrial scenarios. Inspired by the philosophy of "think-like-a-vertex", a GAS-like (Gather-Apply-Scatter) schema is proposed to describe the computation paradigm and data flow of GNN inference. The computation of GNNs is expressed in an iteration manner, in which a vertex would gather messages via in-edges and update its state information by forwarding an associated layer of GNNs with those messages and then send the updated information to other vertexes via out-edges. Following the schema, the proposed InferTurbo can be built with alternative backends (e.g., batch processing system or graph computing system). Moreover, InferTurbo introduces several strategies like shadow-nodes and partial-gather to handle nodes with large degrees for better load balancing. With InferTurbo, GNN inference can be hierarchically conducted over the full graph without sampling and redundant computation. Experimental results demonstrate that our system is robust and efficient for inference tasks over graphs containing some hub nodes with many adjacent edges. Meanwhile, the system gains a remarkable performance compared with the traditional inference pipeline, and it can finish a GNN inference task over a graph with tens of billions of nodes and hundreds of billions of edges within 2 hours.

In this paper, we consider the problem of long tail scenario modeling with budget limitation, i.e., insufficient human resources for model training stage and limited time and computing resources for model inference stage. This problem is widely encountered in various applications, yet has received deficient attention so far. We present an automatic system named ALT to deal with this problem. Several efforts are taken to improve the algorithms used in our system, such as employing various automatic machine learning related techniques, adopting the meta learning philosophy, and proposing an essential budget-limited neural architecture search method, etc. Moreover, to build the system, many optimizations are performed from a systematic perspective, and essential modules are armed, making the system more feasible and efficient. We perform abundant experiments to validate the effectiveness of our system and demonstrate the usefulness of the critical modules in our system. Moreover, online results are provided, which fully verified the efficacy of our system.

In recommendation scenarios, there are two long-standing challenges, i.e., selection bias and data sparsity, which lead to a significant drop in prediction accuracy for both Click-Through Rate (CTR) and post-click Conversion Rate (CVR) tasks. To cope with these issues, existing works emphasize on leveraging Multi-Task Learning (MTL) frameworks (Category 1) or causal debiasing frameworks (Category 2) to incorporate more auxiliary data in the entire exposure/inference space $\mathcal{D}$ or debias the selection bias in the click/training space ${\mathcal{O}}$. However, these two kinds of solutions cannot effectively address the not-missing-at-random problem and debias the selection bias in ${\mathcal{O}}$ to fit the inference in $\mathcal{D}$. To fill the research gaps, we propose a Direct entire-space Causal Multi-Task framework, namely DCMT, for post-click conversion prediction in this paper. Specifically, inspired by users’ decision process of conversion, we propose a new counterfactual mechanism to debias the selection bias in $\mathcal{D}$, which can predict the factual CVR and the counterfactual CVR under the soft constraint of a counterfactual prior knowledge. Extensive experiments demonstrate that our DCMT can improve the state-of-the-art methods by an average of 1.07% in term of CVR AUC on the offline datasets and 0.75% in term of PV-CVR on the online A/B test (the Alipay Search). Such improvements can increase millions of conversions per week in real industrial applications, e.g., the Alipay Search.

This study considers the estimation of conditional causal effects in the presence of unmeasured confounding for a balanced panel with treatment imposed at the last time point. To address this, we combine Difference-in-differences (DiD) and tree-based methods and propose a new identification assumption that allows for the violation of the (conditional) parallel trends assumption adopted by most existing DiD methods. Under this new assumption, we prove partial identifiability of the conditional average treatment effect on the treated group (CATT). Our proposed method estimates CATT through a tree-based causal approach, guided by a novel splitting rule that avoids model misspecification and unnecessary auxiliary parameter estimation. The splitting rule measures both the error of fitting observed data and the violation of conditional parallel trends simultaneously. We also develop an ensemble of multiple trees via gradient boosting to further enhance performance. Experimental results on both synthetic and real-world datasets validate the effectiveness of our proposed method.

Commonsense Question Answering (CQA) aims to answer questions that require human commonsense. Closed-book CQA, as one of the subtasks, requires the model to answer questions without retrieving external knowledge, which emphasizes the importance of the model's problem-solving ability. Most previous methods relied on large-scale pre-trained models to generate question-related knowledge while ignoring the crucial role of skills in the process of answering commonsense questions. Generally, skills refer to the learned ability in performing a specific task or activity, which are derived from knowledge and experience. In this paper, we introduce a new approach named Dynamic Skill-aware Commonsense Question Answering (DSCQA), which transcends the limitations of traditional methods by informing the model about the need for each skill in questions and utilizes skills as a critical driver in CQA process. To be specific, DSCQA first employs commonsense skill extraction module to generate various skill representations. Then, DSCQA utilizes dynamic skill module to generate dynamic skill representations. Finally, in perception and emphasis module, various skills and dynamic skill representations are used to help question-answering process. Experimental results on two publicly available CQA datasets show the effectiveness of our proposed model and the considerable impact of introducing skills.

Linear programming (LP) is arguably the most common optimization problem encountered in practical settings. Important examples include machine learning systems optimization, resource allocation, and other decision-making scenarios. However, even with state-of-the-art (SOTA) solvers, it is extremely challenging to solve large-scale problems arising in industry settings, which could have up to billions of decision variables and require solutions within a time limit to meet business demands. This paper proposes PDAS, a Practical Distributed ADMM System to solve such problems with a variant of the Alternating Direction Method of Multipliers (ADMM) algorithm. PDAS offers user-friendly interfaces and provides near-linear speedup thanks to its high scalability and excellent performance. It also comes with a failover mechanism to ensure the stability of the iterative process. The convergence, feasibility, and optimality of PDAS have been verified on two real-world data-sets, resulting in a 10-4 average relative deviation from Gurobi. Although SOTA solvers do have advantages if only considering the solving time when tested on five small and medium-sized public data-sets, PDAS is more promising after including the modeling time. Moreover, when used to solve large-scale LP problems with up to 109 decision variables and 104 constraints in three real-world scenarios, PDAS achieves at least 2x speedups, well beyond the capabilities of SOTA.

Industrial recommender systems usually need to serve multiple scenarios at the same time. In practice, there are various heterogeneous scenarios, since users frequently engage in scenarios with varying intentions and the items within each scenario typically belong to diverse categories. Existing works of multi-scenario recommendation mainly focus on modeling homogeneous scenarios which have similar data distributions. They equally transfer knowledge to each scenario without considering the diversity of heterogeneous scenarios. In this paper, we argue that the heterogeneity in multi-scenario recommendations is a key problem that needs to be solved. To this end, we propose an industrial framework named Scenario-Aware Model-Agnostic Meta Distillation (SAMD) for the multi-scenario recommendation. SAMD aims to provide scenario-aware and model-agnostic knowledge sharing across heterogeneous scenarios by modeling scenarios' relationship and conducting heterogeneous knowledge distillation. Specifically, SAMD first measures the comprehensive representation of each scenario and then proposes a novel meta distillation paradigm to conduct scenario-aware knowledge sharing. The meta network first establishes the potential scenarios' relationships and generates the strategies of knowledge sharing for each scenario. Then the heterogeneous knowledge distillation utilizes scenario-aware strategies to share knowledge across heterogeneous scenarios through intermediate features distillation without the restriction of the model architecture. In this way, SAMD shares knowledge across heterogeneous scenarios in a scenario-aware and model-agnostic manner, which addresses the problem of heterogeneity. Compared with other state-of-the-art methods, extensive offline experiments, and online A/B testing demonstrate the superior performance of the proposed SAMD framework, especially in heterogeneous scenarios.

User financial default prediction plays a critical role in credit risk forecasting and management. It aims at predicting the probability that the user will fail to make the repayments in the future. Previous methods mainly extract a set of user individual features regarding his own profiles and behaviors and build a binary-classification model to make default predictions. However, these methods cannot get satisfied results, especially for users with limited information. Although recent efforts suggest that default prediction can be improved by social relations, they fail to capture the higher-order topology structure at the level of small subgraph patterns. In this paper, we fill in this gap by proposing a motif-preserving Graph Neural Network with curriculum learning (MotifGNN) to jointly learn the lower-order structures from the original graph and higher-order structures from multi-view motif-based graphs for financial default prediction. Specifically, to solve the problem of weak connectivity in motif-based graphs, we design the motif-based gating mechanism. It utilizes the information learned from the original graph with good connectivity to strengthen the learning of the higher-order structure. And considering that the motif patterns of different samples are highly unbalanced, we propose a curriculum learning mechanism on the whole learning process to more focus on the samples with uncommon motif distributions. Extensive experiments on one public dataset and two industrial datasets all demonstrate the effectiveness of our proposed method.

The recently explosive growth of Super Apps brings great convenience to people's daily life by providing a wide variety of services through mini-programs, including online shopping, travel, finance, and so on. Due to the considerable gap between various scenarios, the restriction of effective information transfer and sharing severely blocks the efficient delivery of online services, potentially affecting the user's app experience. To deeply understand users' needs, we propose SupKG, a commonsense knowledge graph towards Super APP to help comprehensively characterize user behaviors across different business scenarios. In particular, our SupKG is carefully established from multiplex and heterogeneous data source in Alipay (a well-known Super App in China), which also emphasize abundant spatiotemporal relations and intent-related entities to answer the fundamental question in life service ''which service do users need at what time and where''.

This paper proposes a novel strategy for estimating the heterogeneous treatment effect called the Fused and Accurate Shrinkage Tree ($\mathrm{FAST}$). Our approach utilizes both trial and observational data to improve the accuracy and robustness of the estimator. Inspired by the concept of shrinkage estimation in statistics, we develop an optimal weighting scheme and a corresponding estimator that balances the unbiased estimator based on the trial data with the potentially biased estimator based on the observational data. Specifically, combined with tree-based techniques, we introduce a new split criterion that utilizes both trial data and observational data to more accurately estimate the treatment effect. Furthermore, we confirm the consistency of our proposed tree-based estimator and demonstrate the effectiveness of our criterion in reducing prediction error through theoretical analysis. The advantageous finite sample performance of the $\mathrm{FAST}$ and its ensemble version over existing methods is demonstrated via simulations and real data analysis.

Large language models have shown astonishing performance on a wide range of reasoning tasks. In this paper, we investigate whether they could reason about real-world events and help improve the prediction performance of event sequence models. We design LAMP, a framework that integrates a large language model in event prediction. Particularly, the language model performs abductive reasoning to assist an event sequence model: the event model proposes predictions on future events given the past; instructed by a few expert-annotated demonstrations, the language model learns to suggest possible causes for each proposal; a search module finds out the previous events that match the causes; a scoring function learns to examine whether the retrieved events could actually cause the proposal. Through extensive experiments on several challenging real-world datasets, we demonstrate that our framework---thanks to the reasoning capabilities of large language models---could significantly outperform the state-of-the-art event sequence models.

The issue of distribution shifts is emerging as a critical concern in graph representation learning. From the perspective of invariant learning and stable learning, a recently well-established paradigm for out-of-distribution generalization, stable features of the graph are assumed to causally determine labels, while environmental features tend to be unstable and can lead to the two primary types of distribution shifts. The correlation shift is often caused by the spurious correlation between environmental features and labels that differs between the training and test data; the covariate shift often stems from the presence of new environmental features in test data. However, most strategies, such as invariant learning or graph augmentation, typically struggle with limited training environments or perturbed stable features, thus exposing limitations in handling the problem of covariate shift. To address this challenge, we propose a simple-yet-effective data augmentation strategy, Adversarial Invariant Augmentation (AIA), to handle the covariate shift on graphs. Specifically, given the training data, AIA aims to extrapolate and generate new environments, while concurrently preserving the original stable features during the augmentation process. Such a design equips the graph classification model with an enhanced capability to identify stable features in new environments, thereby effectively tackling the covariate shift in data. Extensive experiments with in-depth empirical analysis demonstrate the superiority of our approach. The implementation codes are publicly available at https://github.com/yongduosui/AIA.

Neural Temporal Point Processes (TPPs) are the prevalent paradigm for modeling continuous-time event sequences, such as user activities on the web and financial transactions. In real world applications, the event data typically comes in a streaming manner, where the distribution of the patterns may shift over time. Under the privacy and memory constraints commonly seen in real scenarios, how to continuously monitor a TPP to learn the streaming event sequence is an important yet under-investigated problem. In this work, we approach this problem by adopting Continual Learning (CL), which aims to enable a model to continuously learn a sequence of tasks without catastrophic forgetting. While CL for event sequence is less well studied, we present a simple yet effective framework, PromptTPP, by integrating the base TPP with a continuous-time retrieval prompt pool. In our proposed framework, prompts are small learnable parameters, maintained in a memory space and jointly optimized with the base TPP so that the model is properly instructed to learn event streams arriving sequentially without buffering past examples or task-specific attributes. We formalize a novel and realistic experimental setup for modeling event streams, where PromptTPP consistently sets state-of-the-art performance across two real user behavior datasets.

In marketing recommendations, the campaign organizers will distribute coupons to users to encourage consumption. In general, a series of strategies are employed to interfere with the coupon distribution process, leading to a growing imbalance between user-coupon interactions, resulting in a bias in the estimation of conversion probabilities. We refer to the estimation bias as the matching bias. In this paper, we explore how to alleviate the matching bias from the causal-effect perspective. We regard the historical distributions of users and coupons over each other as confounders and characterize the matching bias as a confounding effect to reveal and eliminate the spurious correlations between user-coupon representations and conversion probabilities. Then we propose a new training paradigm named De-Matching Bias Recommendation (DMBR) to remove the confounding effects during model training via the backdoor adjustment. We instantiate DMBR on two representative models: DNN and MMOE, and conduct extensive offline and online experiments to demonstrate the effectiveness of our proposed paradigm.

In this paper, we highlight that both conformity and risk preference matter in making fund investment decisions beyond personal interest and seek to jointly characterize these aspects in a disentangled manner. Consequently, we develop a novel Multi-granularity Graph Disentangled Learning framework named MGDL to effectively perform intelligent matching of fund investment products. Benefiting from the well-established fund graph and the attention module, multi-granularity user representations are derived from historical behaviors to separately express personal interest, conformity and risk preference in a fine-grained way. To attain stronger disentangled representations with specific semantics, MGDL explicitly involve two self-supervised signals, ie fund type based contrasts and fund popularity. Extensive experiments in offline and online environments verify the effectiveness of MGDL.

Online allocation is a critical challenge in constrained recommendation systems, where the distribution of goods, ads, vouchers, and other content to users with limited resources needs to be managed effectively. While the existing literature has made significant progress in improving recommendation algorithms for various scenarios, less attention has been given to developing and deploying industry-scale online allocation system in an efficient manner. To address this issue, this paper introduces an integrated and efficient learning framework in constrained recommendation scenarios at Alipay. The framework has been tested through experiments, demonstrating its superiority over other state-of-the-art methods.

In industrial recommendation systems, both data sizes and computational resources vary across different scenarios. For scenarios with limited data, data sparsity can lead to a decrease in model performance. Heterogeneous knowledge distillation-based transfer learning can be used to transfer knowledge from models in data-rich domains. However, in recommendation systems, the target domain possesses specific privileged features that significantly contribute to the model. While existing knowledge distillation methods have not taken these features into consideration, leading to suboptimal transfer weights. To overcome this limitation, we propose a novel algorithm called Uncertainty-based Heterogeneous Privileged Knowledge Distillation (UHPKD). Our method aims to quantify the knowledge of both the source and target domains, which represents the uncertainty of the models. This approach allows us to derive transfer weights based on the knowledge gain, which captures the difference in knowledge between the source and target domains. Experiments conducted on both public and industrial datasets demonstrate the superiority of our UHPKD algorithm compared to other state-of-the-art methods.

Meta-learning has become a widely used method for the user cold-start problem in recommendation systems, as it allows the model to learn from similar learning tasks and transfer the knowledge to new tasks. However, most existing meta-learning methods do not consider the temporal factor of users' preferences, which is crucial for news recommendation scenarios where news streams change dynamically over time. In this paper, we propose Time-Aware Meta-Learning (TAML), a novel framework that focuses on cold-start users in news recommendation systems. TAML factorizes user preferences into time-specifc and time-shift representations that jointly affect users' news preferences. These temporal factors are further incorporated into the meta-learning framework to achieve accurate and timely cold-start recommendations. Extensive experiments are conducted on two real-world datasets, demonstrating the superior performance of TAML over state-of-the-art methods.

Transformer-based models have achieved tremendous success in sequential recommendation (SR), but they suffer from consuming excessive computational resources, particularly in the inference stage. Thus, developing lightweight yet effective SR models has become a frequent demand in industrial applications, which is also in line with the ideals of Green AI and Green IR. In this applied paper, we introduce GreenSeq deployed in Alipay to automatically design Green networks that can provide appropriate recommendations with lower computational consumption in SR. Specifically, GreenSeq uses a novel multi-layer search space that allows for flexible network design and a Greenness-aware loss term for balancing efficiency and effectiveness. Experiments on benchmark datasets and A/B testing show that GreenSeq performs well while using fewer resources. GreenSeq also reduces electricity and carbon emissions in Alipay.

The pre-training paradigm, i.e., learning universal knowledge across a wide spectrum of domains, has increasingly become a new de-facto practice in many fields, especially for transferring to new domains. The recent progress includes universal pre-training solutions for recommendation. However, we argue that the common treatment utilizing the masked language modeling or simple data augmentation via contrastive learning is not sufficient for pre-training a recommender system, since a user's intent could be more complex than predicting the next word or item. It is more intuitive to go a step further by devising the multi-interest driven pre-training framework for universal user understanding. Nevertheless, incorporating multi-interest modeling in recommender system pre-training is non-trivial due to the dynamic, contextual, and temporary nature of the user interests, particularly when the users are from different domains. The limited effort on this line has greatly rendered it as an open question.

Aiming at helping users locally discover retail services (e.g., entertainment and dining) on Online to Offline (O2O) service platforms, we propose COUPA, an industrial system targeting for characterizing user preference with inspiring considerations of time and position aware preferences. We carefully implement and deploy COUPA in Alipay with a cooperation of edge, streaming and batch computing, as well as a two-stage online serving mode, to support several popular recommendation scenarios. Extensive experiments reveal the superior performance of COUPA for recommendation.

By treating users' interactions as a user-item graph, graph learning models have been widely deployed in Collaborative Filtering~(CF) based recommendation. Recently, researchers have introduced Graph Contrastive Learning~(GCL) techniques into CF to alleviate the sparse supervision issue, which first constructs contrastive views by data augmentations and then provides self-supervised signals by maximizing the mutual information between contrastive views. Despite the effectiveness, we argue that current GCL-based recommendation models are still limited as current data augmentation techniques, either structure augmentation or feature augmentation. First, structure augmentation randomly dropout nodes or edges, which is easy to destroy the intrinsic nature of the user-item graph. Second, feature augmentation imposes the same scale noise augmentation on each node, which neglects the unique characteristics of nodes on the graph.

In recent years, many methods have been proposed to alleviate the biases in recommender systems by combining biased data and unbiased data. Among these methods, data imputation method is effective, but previous works only employ a straightforward model to generate imputed data, which can not fully characterize the data. In this paper, we propose a novel data imputation approach that combines an unbiased model and a debiasing model with adaptively learnt weights. We conduct extensive experiments on two public recommendation datasets and one production dataset to demonstrate the effectiveness and robustness of the proposed method.

To improve the marketing performance in the financial scenario, it is necessary to develop a trustworthy model to analyze and select promotion-sensitive customers. Bayesian Network (BN) is suitable for this task because of its interpretability and flexibility, but it usually suffers the exponentially growing computation complexity as the number of nodes grows. To tackle this problem, we present a comprehensive distributed platform named DistriBayes, which can efficiently learn, infer and attribute on a large-scale BN all-in-one platform. It implements several score-based structure learning methods, loopy belief propagation with backdoor adjustment for inference, and a carefully optimized search procedure for attribution. Leveraging the distributed cluster, DistriBayes can finish the learning and attribution on Bayesian Network with hundreds of nodes and millions of samples in hours.

Time series forecasting is an important ingredient in the intelligence of business and decision processes. In industrial scenarios, the time series of interest are mostly macroscopic time series that are aggregated from microscopic time series, e.g., the retail sales is aggregated from the sales of different goods, and that are also intervened by certain treatments on the microscopic individuals, e.g., issuing discount coupons on some goods to increase the retail sales. These characteristics are not considered in existing toolkits, which just focus on the "natural" time series forecasting that predicts the future value based on historical data, regardless of the impact of treatments. In this paper, we present AntTS, a time series toolkit paying more attention on the forecasting of the macroscopic time series with underlying microscopic time series and certain treatments, besides the "natural" time series forecasting. AntTS consists of three decoupled modules, namely Clustering module, Natural Forecasting module, and Effect module, which are utilized to study the homogeneous groups of microscopic individuals, the "natural" time series forecasting of homogeneous groups, and the treatment effect estimation of homogeneous groups. With the combinations of different modules, it can exploit the microscopic individuals and the interventions on them, to help the forecasting of macroscopic time series. We show that AntTS helps address many typical tasks in the industry.

In this paper, we present a framework to deal with the fraud detection task with extremely few labeled frauds. We involve human intelligence in the loop in a labor-saving manner and introduce several ingenious designs to the model construction process. Namely, a rule mining module is introduced, and the learned rules will be refined with expert knowledge. The refined rules will be used to relabel the unlabeled samples and get the potential frauds. We further present a model to learn with the reliable frauds, the potential frauds, and the rest normal samples. Note that the label noise problem, class imbalance problem, and confirmation bias problem are all addressed with specific strategies when building the model. Experimental results are reported to demonstrate the effectiveness of the framework.

The power of artificial intelligence (AI) models originates with sophisticated model architecture as well as the sheer size of the model. These large-scale AI models impose new and challenging system requirements regarding scalability, reliability, and flexibility. One of the most promising solutions in the industry is to train these large-scale models on distributed deep-learning frameworks. With the power of all distributed computations, it is desired to achieve a training process with excellent scalability, elastic scheduling (flexibility), and fault tolerance (reliability). In this paper, we demonstrate the scalability, flexibility, and reliability of our open-source Elastic Deep Learning (ElasticDL) framework. Our ElasticDL utilizes an open-source system, i.e., Kubernetes, for automating deployment, scaling, and management of containerized application features to provide fault tolerance and support elastic scheduling for DL tasks.

Collaborative filtering based recommendation learns users’ preferences from all users’ historical behavior data, and has been popular to facilitate decision making. Recently, the fairness issue of recommendation has become more and more essential. A recommender system is considered unfair when it does not perform equally well for different user groups according to users’ sensitive attributes (e.g., gender, race). Plenty of methods have been proposed to alleviate unfairness by optimizing a predefined fairness goal or changing the distribution of unbalanced training data. However, they either suffered from the specific fairness optimization metrics or relied on redesigning the current recommendation architecture. In this paper, we study how to improve recommendation fairness from the data augmentation perspective. The recommendation model amplifies the inherent unfairness of imbalanced training data. We augment imbalanced training data towards balanced data distribution to improve fairness. Given each real original user-item interaction record, we propose the following hypotheses for augmenting the training data: each user in one group has a similar item preference (click or non-click) as the item preference of any user in the remaining group. With these hypotheses, we generate “fake" interaction behaviors to complement the original training data. After that, we design a bi-level optimization target, with the inner optimization generates better fake data to augment training data with our hypotheses, and the outer one updates the recommendation model parameters based on the augmented training data. The proposed framework is generally applicable to any embedding-based recommendation, and does not need to pre-define a fairness metric. Extensive experiments on two real-world datasets clearly demonstrate the superiority of our proposed framework. We publish the source code at https://github.com/newlei/FDA.

A large body of work in Computer Science and Operations Research study online algorithms for stochastic resource allocation problems. The most common assumption is that the online requests have randomly generated i.i.d. types. This assumption is well justified for static markets and/or relatively short time periods. We consider dynamic markets, whose states evolve as a random walk in a market-specific Markov Chain. This is a new model that generalizes previous i.i.d. settings. We identify important parameters of the Markov chain that is crucial for obtaining good approximation guarantees to the expected value of the optimal offline algorithm which knows realizations of all requests in advance. We focus on a stylized single-resource setting and: (i) generalize the well-known Prophet Inequality from the optimal stopping theory (single-unit setting) to Markov Chain setting; (ii) in multi-unit setting, design a simple algorithm that is asymptotically optimal under mild assumptions on the underlying Markov chain.

Uplift modeling aims to measure the incremental effect, which we call uplift, of a strategy or action on the users from randomized experiments or observational data. Most existing uplift methods only use individual data, which are usually not informative enough to capture the unobserved and complex hidden factors regarding the uplift. Furthermore, uplift modeling scenario usually has scarce labeled data, especially for the treatment group, which also poses a great challenge for model training. Considering that the neighbors’ features and the social relationships are very informative to characterize a user’s uplift, we propose a graph neural network-based framework with two uplift estimators, called GNUM, to learn from the social graph for uplift estimation. Specifically, we design the first estimator based on a class-transformed target. The estimator is general for all types of outcomes, and is able to comprehensively model the treatment and control group data together to approach the uplift. When the outcome is discrete, we further design the other uplift estimator based on our defined partial labels, which is able to utilize more labeled data from both the treatment and control groups, to further alleviate the label scarcity problem. Comprehensive experiments on a public dataset and two industrial datasets show a superior performance of our proposed framework over state-of-the-art methods under various evaluation metrics. The proposed algorithms have been deployed online to serve real-world uplift estimation scenarios.

Integrating AI algorithms into databases is an ongoing effort in both academia and industry. We introduce SQLFlow, a toolkit seamlessly combining data manipulations and AI operations that can be run locally or remotely. SQLFlow extends SQL syntax to support typical AI tasks including model training, inference, interpretation, and mathematical optimization. It is compatible with a variety of database management systems (DBMS) and AI engines, including MySQL, TiDB, MaxCompute, and Hive, as well as TensorFlow, scikit-learn, and XGBoost. Documentations and case studies are available at https://sqlflow.org. The source code and additional details can be found at https://github.com/sql-machine-learning/sqlflow.

With the increasing need for graph analysis, massive Concurrent iterative Graph Processing (CGP) jobs are usually performed on the common large-scale real-world graph. Although several solutions have been proposed, these CGP jobs are not coordinated with the consideration of the inherent dependencies in graph data driven by graph topology. As a result, they suffer from redundant and fragmented accesses of the same underlying graph dispersed over distributed platform, because the same graph is typically irregularly traversed by these jobs along different paths at the same time. In this work, we develop GraphTune , which can be integrated into existing distributed graph processing systems, such as D-Galois, Gemini, PowerGraph, and Chaos, to efficiently perform CGP jobs and enhance system throughput. The key component of GraphTune is a dependency-aware synchronous execution engine in conjunction with several optimization strategies based on the constructed cross-iteration dependency graph of chunks. Specifically, GraphTune transparently regularizes the processing behavior of the CGP jobs in a novel synchronous way and assigns the chunks of graph data to be handled by them based on the topological order of the dependency graph so as to maximize the performance. In this way, it can transform the irregular accesses of the chunks into more regular ones so that as many CGP jobs as possible can fully share the data accesses to the common graph. Meanwhile, it also efficiently synchronizes the communications launched by different CGP jobs based on the dependency graph to minimize the communication cost. We integrate it into four cutting-edge distributed graph processing systems and a popular out-of-core graph processing system to demonstrate the efficiency of GraphTune. Experimental results show that GraphTune improves the throughput of CGP jobs by 3.1∼6.2, 3.8∼8.5, 3.5∼10.8, 4.3∼12.4, and 3.8∼6.9 times over D-Galois, Gemini, PowerGraph, Chaos, and GraphChi, respectively.

Vertical federated learning (VFL) is an emerging privacy-preserving paradigm that enables collaboration between companies. These companies have the same set of users but different features. One of them is interested in expanding new business or improving its current service with others’ features. For instance, an e-commerce company, who wants to improve its recommendation performance, can incorporate users’ preferences from another corporation such as a social media company through VFL. On the other hand, graph data is a powerful and sensitive type of data widely used in industry. Their leakage, e.g., the node leakage and/or the relation leakage, can cause severe privacy issues and financial loss. Therefore, protecting the security of graph data is important in practice. Though a line of work has studied how to learn with graph data in VFL, the privacy risks remain underexplored. In this paper, we perform the first systematic study onrelation inference attacksto reveal VFL's risk of leaking samples’ relations. Specifically, we assume the adversary to be a semi-honest participant. Then, according to the adversary's knowledge level, we formulate three kinds of attacks based on different intermediate representations. Particularly, we design a novel numerical approximation method to handle VFL's encryption mechanism on the participant's representations. Extensive evaluations with four real-world datasets demonstrate the effectiveness of our attacks. For instance, the area under curve of relation inference can reach more than 90%, implying an impressive relation inference capability. Furthermore, we evaluate possible defenses to examine our attacks’ robustness. The results show that their impacts are limited. Our work highlights the need for advanced defenses to protect private relations and calls for more exploration of VFL's privacy and security issues.

Online Microlending, a new financial service, focuses on small loans without any sort of collateral. It provides more flexible and quicker funding for borrowers, as well as higher interest rates of return. For platforms that provide such services, an essential task is to adequately evaluate each loan’s risk so as to minimize the possible financial loss. However, there exists a special group of borrowers, namelyfraud-agents, who gain illegal profits from inciting other borrowers to cheat, i.e., they help the high-risk borrowers evade the risk evaluation by crafting fake personal information. The existence of fraud-agents poses a severe threat to the risk management systems and results in a huge financial loss for lending platforms. In this article, we present the first machine learning-based solution to detect fraud-agents in online microlending. The key challenge of this decade-long problem is that it is unclear how to construct effective features from multiple behavior logs such as phone call history, address book, loan history and activity logs of borrowers. To address this problem, we first conduct an empirical study on over 600K borrowers to gain some insights on the adversarial behaviors of fraud-agents comparing to normal borrowers and benign-agents. Based on the study, we are able to design a total of 26 features, falling into four groups, for fraud agent detection. Then, we propose a two-stage detection model to address the challenge of limited number of labeled fraud agent examples. The evaluation results show that our method can achieve a precision of 94.30%. We deploy our method on a real large online microlending platform with 11,953,273 borrowers, and we identify 29,727 fraud-agents from them. The domain experts from the platform confirm that 95.59% of them are real fraud-agents, and have added them to the platform’s internal blacklist. We further conduct a measurement study on those fraud-agents to share deeper insights on their adversarial behaviors.

Aspect-level sentiment analysis is a granular emotional classification task that refers to identifying sentiment polarities towards aspects in a sentence. Although previous research has reached a great achievement, this task remains very challenging. First, previous approaches only focus on one specific domain, which lacks the capability of transferring to other domains. Moreover, the majority of prior studies ignore the direct relationship between aspects and the corresponding sentiment words. To this end, in this paper, we propose a novel model named Efficient Adaptive Transfer Network (EATN) for aspect-level sentiment analysis which emphasizes the need of incorporating the correlation among multiple domains. The proposed EATN provides a Domain Adaptation Module (DAM) to learn common features from the sufficiently labeled source domain and to guide the classification performance in the target domain. Specifically, DAM comprises two special tasks, with one sentiment classification task aiming to learn sentiment knowledge and the other domain classification task focusing on learning domain-invariant features. Moreover, we design aspect-aware multi-head attention mechanism to capture the direct associations between the aspects and the contextual sentiment words, which is beneficial to learn the aspect-aware semantic knowledge. Extensive experiments demonstrate the effectiveness and universality of our method.

Thanks to the inherent spatial or sequential structures underlying the data like images and texts, deep architectures such as convolutional neural networks (CNNs) and the Transformer have been recognized as the preeminent approaches in image processing and language modeling. In the real world, there are a large number of tabular data without any explicit structures, which breaks the inductive bias of most neural networks like CNNs. Although multi-layer perceptrons (MLPs) obtain empirical success on tabular data, they cannot well explain the underlying relationship between multiple variables. Compared with other fields, research on deep models toward tabular data has received relatively less scrutiny. To bridge this gap, we propose Dual-Route Structure-Adaptive Graph Networks (DRSA-Net) to model the nonlinearity in tabular feature vectors without any prior. DRSA-Net adaptively learns a sparse graph structure between variables and then characterizes interactions between them from the view of dual-route message passing. We demonstrate that DRSA-Net could easily degenerate into the typical MLPs and factorization machines (FMs). Extensive experiments on recommendations, images (no spatial information after preprocessing), and some benchmark machine learning datasets show that DRSA-Net achieves comparable or superior performance with many classic algorithms and recently proposed deep models.

Cross-Domain Recommendation (CDR) and Cross-System Recommendation (CSR) have been proposed to improve the recommendation accuracy in a target dataset (domain/system) with the help of a source one with relatively richer information. However, most existing CDR and CSR approaches are single-target, namely, there is a single target dataset, which can only help the target dataset and thus cannot benefit the source dataset. In this paper, we focus on three new scenarios, i.e., Dual-Target CDR (DTCDR), Multi-Target CDR (MTCDR), and CDR+CSR, and aim to improve the recommendation accuracy in all datasets simultaneously for all scenarios. To do this, we propose a unified framework, called GA (based onGraph embedding andAttention techniques), for all three scenarios. In GA, we first construct separate heterogeneous graphs to generate more representative user and item embeddings. Then, we propose an element-wise attention mechanism to effectively combine the embeddings of common entities (users/items) learned from different datasets. Moreover, to avoid negative transfer, we further propose aPersonalized training strategy to minimize the embedding difference of common entities between a richer dataset and a sparser dataset, deriving three new models, i.e., GA-DTCDR-P, GA-MTCDR-P, and GA-CDR+CSR-P, for the three scenarios respectively. Extensive experiments conducted on four real-world datasets demonstrate that our proposed GA models significantly outperform the state-of-the-art approaches.

This paper studies learning node representations with graph neural networks (GNNs) for unsupervised scenario. Specifically, we derive a theoretical analysis and provide an empirical demonstration about the non-steady performance of GNNs over different graph datasets, when the supervision signals are not appropriately defined. The performance of GNNs depends on both the node feature smoothness and the locality of graph structure. To smooth the discrepancy of node proximity measured by graph topology and node feature, we proposed SAIL - a novel self-augmented graph contrastive learning framework, with two complementary self-distilling regularization modules, i.e., intra- and inter-graph knowledge distillation. We demonstrate the competitive performance of SAIL on a variety of graph applications. Even with a single GNN layer, SAIL has consistently competitive or even better performance on various benchmark datasets, comparing with state-of-the-art baselines.

Despite the remarkable performance of graph neural networks (GNNs) in semi-supervised learning, it is criticized for not making full use of unlabeled data and suffering from over-fitting. Recently, graph data augmentation, used to improve both accuracy and generalization of GNNs, has received considerable attentions. However, one fundamental question is how to evaluate the quality of graph augmentations in principle? In this paper, we propose two metrics, Consistency and Diversity, from the aspects of augmentation correctness and generalization. Moreover, we discover that existing augmentations fall into a dilemma between these two metrics. Can we find a graph augmentation satisfying both consistency and diversity? A well-informed answer can help us understand the mechanism behind graph augmentation and improve the performance of GNNs. To tackle this challenge, we analyze two representative semi-supervised learning algorithms: label propagation (LP) and consistency regularization (CR). We find that LP utilizes the prior knowledge of graphs to improve consistency and CR adopts variable augmentations to promote diversity. Based on this discovery, we treat neighbors as augmentations to capture the prior knowledge embodying homophily assumption, which promises a high consistency of augmentations. To further promote diversity, we randomly replace the immediate neighbors of each node with its remote neighbors. After that, a neighbor-constrained regularization is proposed to enforce the predictions of the augmented neighbors to be consistent with each other. Extensive experiments on five real-world graphs validate the superiority of our method in improving the accuracy and generalization of GNNs.

Heterogeneous Graph Neural Networks (HGNNs) have drawn increasing attention in recent years and achieved outstanding performance in many tasks. However, despite their wide use, there is currently no understanding of their robustness to adversarial attacks. In this work, we first systematically study the robustness of HGNNs and show that they can be easily fooled by adding the adversarial edge between the target node and large-degree node (i.e., hub). Furthermore, we show two key reasons for such vulnerability of HGNNs: one is perturbation enlargement effect, i.e., HGNNs, failing to encode transiting probability, will enlarge the effect of the adversarial hub in comparison of GCNs, and the other is soft attention mechanism, i.e., such mechanism assigns positive attention values to obviously unreliable neighbors. Based on the two facts, we propose a novel robust HGNN framework RoHe against topology adversarial attacks by equipping an attention purifier, which can prune malicious neighbors based on topology and feature. Specifically, to eliminate the perturbation enlargement, we introduce the metapath-based transiting probability as the prior criterion of the purifier, restraining the confidence of malicious neighbors from the adversarial hub. Then the purifier learns to mask out neighbors with low confidence, thus can effectively alleviate the negative effect of malicious neighbors in the soft attention mechanism. Extensive experiments on different benchmark datasets for multiple HGNNs are conducted, where the considerable improvement of HGNNs under adversarial attacks will demonstrate the effectiveness and generalization ability of our defense framework.

Pairwise ranking models have been widely used to address recommendation problems. The basic idea is to learn the rank of users' preferred items through separating items into positive samples if user-item interactions exist, and negative samples otherwise. Due to the limited number of observable interactions, pairwise ranking models face serious class-imbalance issues. Our theoretical analysis shows that current sampling-based methods cause the vertex-level imbalance problem, which makes the norm of learned item embeddings towards infinite after a certain training iterations, and consequently results in vanishing gradient and affects the model inference results. We thus propose an efficient Vital Negative Sampler (VINS) to alleviate the class-imbalance issue for pairwise ranking model, in particular for deep learning models optimized by gradient methods. The core of VINS is a bias sampler with reject probability that will tend to accept a negative candidate with a larger degree weight than the given positive item. Evaluation results on several real datasets demonstrate that the proposed sampling method speeds up the training procedure 30% to 50% for ranking models ranging from shallow to deep, while maintaining and even improving the quality of ranking results in top-N item recommendations.

Data augmentation-based semi-supervised learning (SSL) methods have made great progress in computer vision and natural language processing areas. One of the most important factors is that the semantic structure invariance of these data allows the augmentation procedure (e.g., rotating images or masking words) to thoroughly utilize the enormous amount of unlabeled data. However, the tabular data does not possess an obvious invariant structure, and therefore similar data augmentation methods do not apply to it. To fill this gap, we present a simple yet efficient data augmentation method particular designed for tabular data and apply it to the SSL algorithm: SDAT (Semi-supervised learning with Data Augmentation for Tabular data). We adopt a multi-task learning framework that consists of two components: the data augmentation procedure and the consistency training procedure. The data augmentation procedure which perturbs in latent space employs a variational auto-encoder (VAE) to generate the reconstructed samples as augmented samples. The consistency training procedure constrains the predictions to be invariant between the augmented samples and the corresponding original samples. By sharing a representation network (encoder), we jointly train the two components to improve effectiveness and efficiency. Extensive experimental studies validate the effectiveness of the proposed method on the tabular datasets.

Current Graph Neural Networks (GNNs) suffer from the over-smoothing problem, which results in indistinguishable node representations and low model performance with more GNN layers. Many methods have been put forward to tackle this problem in recent years. However, existing tackling over-smoothing methods emphasize model performance and neglect the over-smoothness of node representations. Additional, different approaches are applied one at a time, while there lacks an overall framework to jointly leverage multiple solutions to the over-smoothing challenge. To solve these problems, we propose GraTO, a framework based on neural architecture search to automatically search for GNNs architecture. GraTO adopts a novel loss function to facilitate striking a balance between model performance and representation smoothness. In addition to existing methods, our search space also includes DropAttribute, a novel scheme for alleviating the over-smoothing challenge, to fully leverage diverse solutions. We conduct extensive experiments on six real-world datasets to evaluate GraTo, which demonstrates that GraTo outperforms baselines in the over-smoothing metrics and achieves competitive performance in accuracy. GraTO is especially effective and robust with increasing numbers of GNN layers. Further experiments bear out the quality of node representations learned with GraTO and the effectiveness of model architecture. We make the code of GraTo available at Github (https://github.com/fxsxjtu/GraTO).

Post-processing is crucial to modern recommendation systems to achieve various purposes, e.g., improving diversity, and giving reasonable itineraries which consist of combinations of items, but is merely studied in the literature. We decouple the recommendation system into two modules including a reward estimation module and a post-processing module. Our real-time post-processing module built on Ray abstracts the common post-processing problems in the itinerary recommendation as combinatorial optimization problems. Under the goal of maximizing the click-through rate, the more reasonable recommendation results are obtained by imposing various constraints on the candidate items. However, the optimization problems are typically mixed integer programming problems with quadratic terms in practice, which are NP-hard. In real-time scenarios, there are extremely high requirements for the speed of the solving process. We speed up the problem solving by linearizing and relaxing the original problem and use Ray serving as the underlying service to provide stable and efficient technical support. At last, We provide services to users by deploying the post-processing module in the itinerary recommendation scenario at Alipay's built-in applet named ''What's nearby''. The online A/B experiment shows that the user exposure click rate can be significantly improved.

Decision rules have been successfully used in various classification applications because of their interpretability and efficiency. In many real-world scenarios, especially in industrial applications, it is necessary to generate rule sets under certain constraints, such as confidence constraints. However, most previous rule mining methods only emphasize the accuracy of the rule set but take no consideration of these constraints. In this paper, we propose a Confidence-constraint Rule Set Learning (CRSL) framework consisting of three main components, i.e. rule miner, rule ranker, and rule subset selector. Our method not only considers the trade-off between confidence and coverage of the rule set but also considers the trade-off between interpretability and performance. Experiments on benchmark data and large-scale industrial data demonstrate that the proposed method is able to achieve better performance (6.7% and 8.8% improvements) and competitive interpretability when compared with other rule set learning methods.

Modeling users' historical behaviors is an essential task in many industrial recommender systems. The user interest representation, in previous works, is obtained through the following paradigm: concrete behaviors are firstly embedded as low-dimensional behavior representations, which are then aggregated conditioning on the target item for final user interest representation. Most existing researches focus on the aggregation process that explores the intrinsic structure of the behavior sequences. However, the quality of behavior representation is largely ignored. In this paper, we present a pluggable module, FwSeqBlock, to enhance the expressiveness of behavior representations. Specifically, FwSeqBlock introduces the multiplicative operation among users' historical behaviors and the target item, where a field memory unit is designed to dynamically identify the dominant features from the behavior sequence and filter out the noise. Extensive experiments validate that FwSeqBlock consistently generates higher-quality user representations compared with competitive methods. Besides, online A/B testing reports a 4.46% improvement in Click-Through Rate (CTR), confirming the effectiveness of the proposed method.

In recommender systems, content-based methods and meta-learning involved methods usually have been adopted to alleviate the item cold-start problem. The former consider utilizing item attributes at the feature level and the latter aim at learning a globally shared initialization for all tasks to achieve fast adaptation with limited data at the task level. However, content-based methods only focus on the similarity of item attributes, ignoring the relationships established by user interactions. And for tasks with different distributions, most meta-learning-based methods are difficult to achieve better performance under a single initialization. To address the limitations mentioned above and combine the strengths of both methods, we propose a Task Similarity Aware Meta-Learning (TSAML) framework from two aspects. Specifically, at the feature level, we simultaneously introduce content information and user-item relationships to exploit task similarity. At the task level, we design an automatic soft clustering module to cluster similar tasks and generate the same initialization for similar tasks. Extensive offline experiments demonstrate that the TSAML framework has superior performance and recommends cold items to preferred users more effectively than other state-of-the-art methods.

Machine learning methods for fraud detection have achieved impressive prediction performance, but often sacrifice critical interpretability in many applications. In this work, we propose to learn interpretable models for fraud detection as a simple rule set. More specifically, we design a novel neural rule learning method by building a condition graph with an expectation to capture the high-order feature interactions. Each path in this condition graph can be regarded as a single rule. Inspired by the key idea of meta learning, we combine the neural rules with rules extracted from the tree-based models in order to provide generalizable rule candidates. Finally, we propose a flexible rule set learning framework by designing a greedy optimization method towards maximizing the recall number of fraud samples with a predefined criterion as the cost. We conduct comprehensive experiments on large-scale industrial datasets. Interestingly, we find that the neural rules and rules extracted from tree-based models can be complementary to each other to improve the prediction performance.

Assigning items to owners is a common problem found in various real-world applications, for example, audience-channel matching in marketing campaigns, borrower-lender matching in loan management, and shopper-merchant matching in e-commerce. Given an objective and multiple constraints, an assignment problem can be formulated as a constrained optimization problem. Such assignment problems are usually NP-hard [21], so when the number of items or the number of owners is large, solving for exact solutions becomes challenging. In this paper, we are interested in solving constrained assignment problems with hundreds of millions of items. Thus, with just tens of owners, the number of decision variables is at billion-scale. This scale is usually seen in the internet industry, which makes decisions for large groups of users. We relax the possible integer constraint, and formulate a general optimization problem that covers commonly seen assignment problems. Its objective function is convex. Its constraints are either linear, or convex and separable by items. We study to solve our generalized assignment problems in the Bregman Alternating Direction Method of Multipliers (BADMM) framework where we exploit Bregman divergence to transform the Augmented Lagrangian into a separable form, and solve many subproblems in parallel. The entire solution can thus be implemented using a MapReduce-style distributed computation framework. We present experiment results on both synthetic and real-world datasets to verify its accuracy and scalability.

Credit risk assessment, which aims at identifying high-risk users, plays a critical role in financial institutions. A common method is to use the greedy strategy to generate an interpretable rule set to classify all the users into high-risk or non-risk users. During each iteration, the greedy strategy utilizes a pre-defined indicator function to evaluate which rule is the best and then adds it to the rule set. However, in reality, the indicator function is designed manually and requires much domain knowledge and expert experience. Worse still, we need to design a suitable indicator for every situation, which is tedious and time-consuming work. This motivates us to propose a self-adaptive indicator that can be adapted to different situations without too much human intervention. In this paper, we see the indicator as a weighted sum of several sub-indicators. By tuning the weights, the indicator can be adapted to different situations automatically. That is, we transform this indicator selection problem into a weights tuning problem. To find the best weight of self-adaptive indicators, machine learning methods and black-box optimization are utilized. The experimental results demonstrated that our self-adaptive indicator can select a better rule set to identify more high-risk users compared to the human-defined indicator.

Deep learning based face swapping technologies have opened new frontiers for entertainment industries while pose novel threats to identity security. Applying face swapping to real-world products, as well as defending against its misuse, rely on the capacity to generate high quality face swapped images from realistic scenarios where high resolution images are hard to come by. To this end, we need to address the drawbacks of existing methods, especially on their lacking on maintaining the detail attributes and their dependency on high resolution images as inputs. In this paper, we propose a novel Attribute-Conditioned Face Swapping Network (AFSNet) to preserve attributes and handle low resolution images. Specifically, we use an Image Enhancement Network (IEN) to restore high resolution images from low resolution images and a Face Exchange Module (FEM) to swap the faces. In the FEM, we improve the fidelities of the generated images by using a novel multi-domain feature fusion module (MDFFM) to integrate the identity feature, context feature, IEN feature, and attribute vector to obtain the final image. We also design an attribute transfer loss to promote the consistency of the attributes such as beards and youth between the source and swapped images. The experiments demonstrate our method’s superior performance compared with the state-of-the-art methods.

In this paper, we study the digital marketing where marketing officers (MOs) have to commit to creating brand new promotion ads/contents based on understandings of users' needs or preferences. Users' behaviors are typically high dimensional and hard to understand. Therefore, dimension reduction of users' behaviors from high dimensions and explainability are important to help MOs launch operation-friendly marketings. As such, it is natural to exploit topic models to help MOs understand users' intents from users' behaviors (e.g., user-item visits) in case we treat each user as a document and users' behaviors of visiting an item as a word. However, users of low activities and items followed by power law distributions are common in user-item visit data, which pose significant challenges to traditional topic models. We present a social and semantic enhanced topic model (S2TM) for users' intent mining. We optimize the user-intent estimates based on a graph neural network atop of a social network, and optimize the intent-item estimates based on a skip-gram word embedding approach by linking the semantics of items to pre-trained word embeddings. We propose an efficient stochastic vari-ational inference algorithm for the inference of latent variables and learning of parameters. Extensive experiments on real-world data show the effectivenesses of S2TM in terms of perplexities, topic coherence and semantic coherence compared with state-of-the-art topic models. We further show how MOs interact with our operation-friendly intent mining system, and results on real-world marketing campaigns in terms of click-through rate at Alipay.

E-commerce has gone a long way in empowering merchants through the internet. In order to store the goods efficiently and arrange the marketing resource properly, it is important for them to make the accurate gross merchandise value (GMV) prediction. However, it's nontrivial to make accurate prediction with the deficiency of digitized data. In this article, we present a solution to better forecast GMV inside Alipay app. Thanks to graph neural networks (G NN) which has great ability to correlate different entities to enrich information, we propose Gaia, a graph neural network (GNN) model with temporal shift aware attention. Gaia leverages the relevant e-seller’ sales information and learn neighbor correlation based on temporal dependencies. By testing on Alipay's real dataset and comparing with other baselines, Gaia has shown the best performance. And Gaia is deployed in the simulated online environment, which also achieves great improvement compared with baselines.

Recently, Graph Neural Network (GNN) has achieved remarkable progresses in various real-world tasks on graph data, consisting of node features and the adjacent information between different nodes. High-performance GNN models always depend on both rich features and complete edge information in graph. However, such information could possibly be isolated by different data holders in practice, which is the so-called data isolation problem. To solve this problem, in this paper, we propose VFGNN, a federated GNN learning paradigm for privacy-preserving node classification task under data vertically partitioned setting, which can be generalized to existing GNN models. Specifically, we split the computation graph into two parts. We leave the private data (i.e., features, edges, and labels) related computations on data holders, and delegate the rest of computations to a semi-honest server. We also propose to apply differential privacy to prevent potential information leakage from the server. We conduct experiments on three benchmarks and the results demonstrate the effectiveness of VFGNN.

Recently, representation learning on temporal graphs has drawn increasing attention, which aims at learning temporal patterns to characterize the evolving nature of dynamic graphs in real-world applications. Despite effectiveness, these methods commonly ignore the individual- and combinatorial-level patterns derived from different types of interactions (e.g.,user-item), which are at the heart of the representation learning on temporal graphs. To fill this gap, we propose MERIT, a novel multi-level graph attention network for inductive representation learning on temporal graphs.We adaptively embed the original timestamps to a higher, continuous dimensional space for learn-ing individual-level periodicity through Personalized Time Encoding (PTE) module. Furthermore, we equip MERIT with Continuous time and Con-text aware Attention (Coco-Attention) mechanism which chronologically locates most relevant neighbors by jointly capturing multi-level context on temporal graphs. Finally, MERIT performs multiple aggregations and propagations to explore and exploit high-order structural information for down-stream tasks. Extensive experiments on four public datasets demonstrate the effectiveness of MERITon both (inductive / transductive) link prediction and node classification task.

Multi-dimensional event sequence clustering applies to many scenarios e.g. e-Commerce and electronic health. Traditional clustering models fail to characterize complex real-world processes due to the strong parametric assumption. While Neural Temporal Point Processes (NTPPs) mainly focus on modeling similar sequences instead of clustering. To fill the gap, we propose Mixture of Neural Temporal Point Processes (NTPP-MIX), a general framework that can utilize many existing NTPPs for multi-dimensional event sequence clustering. In NTPP-MIX, the prior distribution of coefficients for cluster assignment is modeled by a Dirichlet distribution. When the assignment is given, the conditional probability of a sequence is modeled by the mixture of a series of NTPPs. We combine variational EM algorithm and Stochastic Gradient Descent (SGD) to efficiently train the framework. In E-step, we fix parameters for NTPPs and approximate the true posterior with variational distributions. In M-step, we fix variational distributions and use SGD to update parameters of NTPPs. Extensive experimental results on four synthetic datasets and three real-world datasets show the effectiveness of NTPP-MIX against state-of-the-arts.

Offline user identification is a scenario that users use their bio-information like faces as identification in offline venues, which has been applied in many offline scenarios such as verification in banks, check-in in hotels and making a purchase in offline merchants. In such a scenario, designing an identification approach to do extremely accurate offline user identification is critical. Most scenarios use faces to identify users and previous algorithms are mainly based on visual features and computer-vision models. However, due to the large variations such as pose, illumination and occlusions in offline scenarios, it remains a challenging problem for existing computer-vision algorithms to get a satisfying accuracy in real-world scenarios. Furthermore, billion-scale candidate users also require high efficiency and high accuracy for the approach.

Unmeasured confounding poses a significant threat to the validity of causal inference. Despite that various ad hoc methods are developed to remove confounding effects, they are subject to certain fairly strong assumptions. In this work, we consider the estimation of conditional causal effects in the presence of unmeasured confounding using observational data and historical controls. Under an interpretable transportability condition, we prove the partial identifiability of conditional average treatment effect on the treated group (CATT). For tree-based models, a new notion, \emph{confounding entropy}, is proposed to measure the discrepancy introduced by unobserved confounders between the conditional outcome distribution of the treated and control groups. The confounding entropy generalizes conventional confounding bias, and can be estimated effectively using historical controls. We develop a new method, debiased causal tree, whose splitting rule is to minimize the empirical risk regularized by the confounding entropy. Notably, our method integrates current observational data (for empirical risk) and their historical controls (for confounding entropy) harmoniously. We highlight that, debiased causal tree can not only estimate CATT well in the presence of unmeasured confounding, but also is a robust estimator of conditional average treatment effect (CATE) against the imbalance of the treated and control populations when all confounders are observed. An extension of combining multiple debiased causal trees to further reduce biases by gradient boosting is considered. The computational feasibility and statistical power of our method are evidenced by simulations and a study of a credit card balance dataset.

Recently, graph neural networks (GNNs) have been shown powerful capacity at modeling structural data. However, when adapted to downstream tasks, it usually requires abundant task-specific labeled data, which can be extremely scarce in practice. A promising solution to data scarcity is to pre-train a transferable and expressive GNN model on large amounts of unlabeled graphs or coarse-grained labeled graphs. Then the pre-trained GNN is fine-tuned on downstream datasets with task-specific fine-grained labels. In this paper, we present a novel Graph Matching based GNN Pre-Training framework, called GMPT. Focusing on a pair of graphs, we propose to learn structural correspondences between them via neural graph matching, consisting of both intra-graph message passing and inter-graph message passing. In this way, we can learn adaptive representations for a given graph when paired with different graphs, and both node- and graph-level characteristics are naturally considered in a single pre-training task. The proposed method can be applied to fully self-supervised pre-training and coarse-grained supervised pre-training. We further propose an approximate contrastive training strategy to significantly reduce time/memory consumption. Extensive experiments on multi-domain, out-of-distribution benchmarks have demonstrated the effectiveness of our approach. The code is available at: https://github.com/RUCAIBox/GMPT.

There exists the cold-start problem in the recommendation systems when observed user-item interactions are insufficient. To alleviate this problem, most existing works aim to learn globally shared prior knowledge across all items and be fast adapted to a new item with few interactions. However, such learning techniques are data demanding and work poorly on new items with no interactions. In this applied paper, we present an industrial framework recently deployed on Alipay to address the item cold-start problem in zero-shot scenarios. The proposed framework provides both efficient and high-quality recommendations for cold items with no log data. Specifically, we formulate the cold-start problem as a zero-shot learning problem and build a highly efficient infrastructure to accomplish online zero-shot recommendations used on large-scale platforms. Extensive offline experiments and online A/B testing demonstrate that the proposed framework has superior performance and recommends cold items to preferred users more effectively than other state-of-the-art methods.

Cold-start diagnosis prediction is a challenging task for AI in healthcare, where often only a few visits per patient and a few observations per disease can be exploited. Although meta-learning is widely adopted to address the data sparsity problem in general domains, directly applying it to healthcare data is less effective, since it is unclear how to capture both the temporal relations in clinical visits and the complicated relations among syndromic diseases for precise personalized diagnosis. To this end, we first propose a novel Meta-learning framework for cold-start diagnosis prediction in healthCare data (MetaCare). By explicitly encoding the effects of disease progress over time as a generalization prior, MetaCare dynamically predicts future diagnosis and timestamp for infrequent patients. Then, to model complicated relations among rare diseases, we propose to utilize domain knowledge of hierarchical relations among diseases, and further perform diagnosis subtyping to mine the latent syndromic relations among diseases. Finally, to tailor the generic meta-learning framework with personalized parameters, we design a hierarchical patient subtyping mechanism and bridge the modeling of both infrequent patients and rare diseases. We term the joint model as MetaCare++. Extensive experiments on two real-world benchmark datasets show significant performance gains brought by MetaCare++, yielding average improvements of 7.71% for diagnosis prediction and 13.94% for diagnosis time prediction over the state-of-the-art baselines.

One intriguing property of adversarial attacks is their “transferability” – an adversarial example crafted with respect to one deep neural network (DNN) model is often found effective against other DNNs as well. Intensive research has been conducted on this phenomenon under simplistic controlled conditions. Yet, thus far there is still a lack of comprehensive understanding about transferability-based attacks (“transfer attacks”) in real-world environments.To bridge this critical gap, we conduct the first large-scale systematic empirical study of transfer attacks against major cloud-based MLaaS platforms, taking the components of a real transfer attack into account. The study leads to a number of interesting findings which are inconsistent to the existing ones, including: (i) Simple surrogates do not necessarily improve real transfer attacks. (ii) No dominant surrogate architecture is found in real transfer attacks. (iii) It is the gap between posterior (output of the softmax layer) rather than the gap between logit (so-called κ value) that increases transferability. Moreover, by comparing with prior works, we demonstrate that transfer attacks possess many previously unknown properties in real-world environments, such as (i) Model similarity is not a well-defined concept. (ii) L2norm of perturbation can generate high transferability without usage of gradient and is a more powerful source than L∞norm. We believe this work sheds light on the vulnerabilities of popular MLaaS platforms and points to a few promising research directions.1

Modern recommendation systems introduce the re-ranking stage to optimize the entire list directly. This paper focuses on the design of re-ranking framework in feed to optimally model the mutual influence between items and further promote user engagement. On mobile devices, users browse the feed almost in a top-down manner and rarely compare items back and forth. Besides, users often compare item with its adjacency based on their partial observations. Given the distinct user behavior patterns, the modeling of mutual influence between items should be carefully designed. Existing re-ranking models encode the mutual influence between items with sequential encoding methods. However, previous works may be dissatisfactory due to the ignorance of connections between items on different scopes. In this paper, we first discuss Unidirectivity and Locality on the impacts and consequences, then report corresponding solutions in industrial applications. We propose a novel framework based on the empirical evidence from user analysis. To address the above problems, we design a \underlineS cope-aware \underlineR e-ranking with \underlineG ated \underlineA ttention model (SRGA ) to emulate the user behavior patterns from two aspects: 1) we emphasize the influence along the user's common browsing direction; 2) we strength the impacts of pivotal adjacent items within the user visual window. Specifically, we design a global scope attention to encode inter-item patterns unidirectionally from top to bottom. Besides, we devise a local scope attention sliding over the recommendation list to underline interactions among neighboring items. Furthermore, we design a learned gate mechanism to aggregating the information dynamically from local and global scope attention. Extensive offline experiments and online A/B testing demonstrate the benefits of our novel framework. The proposed SRGA model achieves the best performance in offline metrics compared with the state-of-the-art re-ranking methods. Further, empirical results on live traffic validate that our recommender system, equipped with SRGA in the re-ranking stage, improves significantly in user engagement.

Graph Convolutional Networks (GCNs) have recently attracted vast interest and achieved state-of-the-art performance on graphs, but its success could typically hinge on careful training with amounts of expensive and time-consuming labeled data. To alleviate labeled data scarcity, self-training methods have been widely adopted on graphs by labeling high-confidence unlabeled nodes and then adding them to the training step. In this line, we empirically make a thorough study for current self-training methods on graphs. Surprisingly, we find that high-confidence unlabeled nodes are not always useful, and even introduce the distribution shift issue between the original labeled dataset and the augmented dataset by self-training, severely hindering the capability of self-training on graphs. To this end, in this paper, we propose a novel Distribution Recovered Graph Self-Training framework (DR-GST), which could recover the distribution of the original labeled dataset. Specifically, we first prove the equality of loss function in self-training framework under the distribution shift case and the population distribution if each pseudo-labeled node is weighted by a proper coefficient. Considering the intractability of the coefficient, we then propose to replace the coefficient with the information gain after observing the same changing trend between them, where information gain is respectively estimated via both dropout variational inference and dropedge variational inference in DR-GST. However, such a weighted loss function will enlarge the impact of incorrect pseudo labels. As a result, we apply the loss correction method to improve the quality of pseudo labels. Both our theoretical analysis and extensive experiments on five benchmark datasets demonstrate the effectiveness of the proposed DR-GST, as well as each well-designed component in DR-GST.

In recent years, machine learning models have achieved magnificent success in many industrial applications, but most of them are black boxes. It is crucial to understand why such predictions are made in many critical areas such as medicine, financial markets, and auto driving. In this paper, we propose Coco, a novel interpretation method which can interpret any binary classifier by assigning each feature an importance value for a particular prediction. We first adopt MixUp method to generate reasonable perturbations, then apply these perturbations with constraints to obtain counterfactual instances and finally compute a comprehensive metric on these instances to estimate the importance of each feature. To demonstrate the effectiveness of Coco, we conduct extensive experiments on several datasets. The results show our method achieves better performance in identifying the most important features compared with the state-of-the-art interpretation methods, including Shap and Lime.

Artificial intelligence (AI) has accelerated the advancement of financial services by identifying hidden patterns from data to improve the quality of financial decisions. However, in addition to commonly desired attributes, such as model accuracy, financial services demand trustworthy AI with properties that have not been adequately realized. These properties of trustworthy AI are interpretability, fairness and inclusiveness, robustness and security, and privacy protection. Here, we review the recent progress and limitations of applying AI to various areas of financial services, including risk management, fraud detection, wealth management, personalized services, and regulatory technology. Based on these progress and limitations, we introduce FinBrain 2.0, a research framework toward trustworthy AI. We argue that we are still a long way from having a truly trustworthy AI in financial services and call for the communities of AI and financial industry to join in this effort.

Deep Neural Networks (DNNs) have achieved remarkable progress in various real-world applications, especially when abundant training data are provided. However, data isolation has become a serious problem currently. Existing works build privacy-preserving DNN models from either algorithmic perspective or cryptographic perspective. The former mainly splits the DNN computation graph between data holders or between data holders and server, which demonstrates good scalability but suffers from accuracy loss and potential privacy risks. In contrast, the latter leverages time-consuming cryptographic techniques, which has strong privacy guarantee but poor scalability. In this article, we propose SPNN—a Scalable and Privacy-preserving deep Neural Network learning framework, from an algorithmic-cryptographic co-perspective. From algorithmic perspective, we split the computation graph of DNN models into two parts, i.e., the private-data-related computations that are performed by data holders and the rest heavy computations that are delegated to a semi-honest server with high computation ability. From cryptographic perspective, we propose using two types of cryptographic techniques, i.e., secret sharing and homomorphic encryption, for the isolated data holders to conduct private-data-related computations privately and cooperatively. Furthermore, we implement SPNN in a decentralized setting and introduce user-friendly APIs. Experimental results conducted on real-world datasets demonstrate the superiority of our proposed SPNN.

Privacy-preserving machine learning has drawn increasingly attention recently, especially with kinds of privacy regulations come into force. Under such situation, Federated Learning (FL) appears to facilitate privacy-preserving joint modeling among multiple parties. Although many federated algorithms have been extensively studied, there is still a lack of secure and practical gradient tree boosting models (e.g., XGB) in literature. In this paper, we aim to build large-scale secure XGB under vertically federated learning setting. We guarantee data privacy from three aspects. Specifically, (1) we employ secure multi-party computation techniques to avoid leaking intermediate information during training, (2) we store the output model in a distributed manner in order to minimize information release, and (3) we provide a novel algorithm for secure XGB predict with the distributed model. Furthermore, by proposing secure permutation protocols, we can improve the training efficiency and make the framework scale to large dataset. We conduct extensive experiments on both public datasets and real-world datasets, and the results demonstrate that our proposed XGB models provide not only competitive accuracy but also practical performance.

Understanding inactive users is the key to user growth and engagement for many Internet companies. However, learning inactive users' representations and their preferences is still challenging because the features available are missing and the positive responses or labels are insufficient. In this paper, we propose a cross domain learning approach to exclusively recommend customized items to inactive users by leveraging the knowledge of active users. Particularly, we represent users, no matter active or inactive users, by their friends' browsing behaviors based on a graph neural network (GNN) layer atop of a heterogeneous graph defined on social networks (user-user friendships) and browsing behaviors (user-page clicks). We jointly optimize the learning tasks of active users in source domain and inactive users in target domain based on the domain invariant features extracted from the embedding of our GNN layer, where the domain invariant features that are learned to benefit both tasks on active/inactive users, and are indiscriminate with respect to the shift between the domains. Extensive experiments show that our approach can well capture the preference of inactive users using both public data and real-world data at Alipay.

Knowledge graph is generally incorporated into recommender systems to improve overall performance. Due to the generalization and scale of the knowledge graph, most knowledge relationships are not helpful for a target user-item prediction. To exploit the knowledge graph to capture target-specific knowledge relationships in recommender systems, we need to distill the knowledge graph to reserve the useful information and refine the knowledge to capture the users' preferences. To address the issues, we propose Knowledge-aware Conditional Attention Networks (KCAN), which is an end-to-end model to incorporate knowledge graph into a recommender system. Specifically, we use a knowledge-aware attention propagation manner to obtain the node representation first, which captures the global semantic similarity on the user-item network and the knowledge graph. Then given a target, i.e., a user-item pair, we automatically distill the knowledge graph into the target-specific subgraph based on the knowledge-aware attention. Afterward, by applying a conditional attention aggregation on the subgraph, we refine the knowledge graph to obtain target-specific node representations. Therefore, we can gain both representability and personalization to achieve overall performance. Experimental results on real-world datasets demonstrate the effectiveness of our framework over the state-of-the-art algorithms.

Cold start problem is one of the most challenging and long-standing problems in recommender systems, and cross-domain recommendation (CDR) methods are effective for tackling it. Most cold-start related CDR methods require training a mapping function between high-dimensional embedding space using overlapping user data. However, the overlapping data is scarce in many recommendation tasks, which makes it difficult to train the mapping function. In this paper, we propose a new approach for CDR, which aims to alleviate the training difficulty. The proposed method can be viewed as a special parameterization of the mapping function without hurting expressiveness, which makes use of non-overlapping user data and leads to effective optimization. Extensive experiments on two real-world CDR tasks are performed to evaluate the proposed method. In the case that there are few overlapping data, the proposed method outperforms the existed state-of-the-art method by 14% (relative improvement).

Incorporating review information into the recommender system has been demonstrated to be an effective method for boosting the recommendation performance. Previous research mainly focus on designing advanced architectures to better profile the users and items. However, the review information in realities can be highly sparse and imbalanced, which poses great challenges for effective user/item representations and satisfied performance enhancement. To alleviate this problem, in this paper, we propose to improve review-based recommendation by counterfactually augmenting the training samples. We focus on a common setting --- feature-aware recommendation, and the main building block of our idea lies in the counterfactual question: "what would be the user's decision if her feature-level preference had been different?''. When augmenting the training samples, we actively change the user preference (also called intervention), and predict the user feedback on the items based on pre-trained recommender models. Instead of changing the user preference in a random manner, we design a learning-based method to discover the samples which are more effective for model optimization. In order to improve the sample qualities, we propose two strategies --- constrained feature perturbation and frequency-based sampling --- to equip our model. Since the sample generation model can be not perfect, we theoretically analyze the relation between the model prediction error and the number of generated samples. As a byproduct, our framework can explain the user pair-wise preference, which is complementary to the traditional point-wise explanations. Extensive experiments demonstrate that our model can significantly improve the performance of the state-of-the-art methods.

Recent studies in recommender systems have managed to achieve significantly improved performance. However, despite being extensively studied, these methods still suffer from two limitations. First, previous studies either encode the document or extract latent sentiment via neural networks, which are difficult to interpret the sentiment of reviewers intuitively. Second, they neglect the personalized interaction of reviews with user/item, i.e., each review has different contributions when modeling the preference of user/item

The orderly operation and development of any system are indivisible from decision optimization. Several issues in life are applicable to the thought of optimization problems to resolve. In this digital age, the size of information and data is obtaining larger and the potency of problem determination is changing into more demanding. Though there're many solvers for specific optimization problems, in the face of large-scale scenarios, there's no single platform that concurrently addresses usability, solvers' uniformity, and computing efficiency. In this demo, we present AntOpt, a decision optimization platform that integrates large-scale distributed computing engines, optimization algorithm solvers and productized services.

Face recognition is greatly improved by deep convolutional neural networks (CNNs). Recently, these face recognition models have been used for identity authentication in security sensitive applications. However, deep CNNs are vulnerable to adversarial patches, which are physically realizable and stealthy, raising new security concerns on the real-world applications of these models. In this paper, we evaluate the robustness of face recognition models using adversarial patches based on transferability, where the attacker has limited accessibility to the target models. First, we extend the existing transfer-based attack techniques to generate transferable adversarial patches. However, we observe that the transferability is sensitive to initialization and degrades when the perturbation magnitude is large, indicating the overfitting to the substitute models. Second, we propose to regularize the adversarial patches on the low dimensional data manifold. The manifold is represented by generative models pre-trained on legitimate human face images. Using face-like features as adversarial perturbations through optimization on the manifold, we show that the gaps between the responses of substitute models and the target models dramatically decrease, exhibiting a better transferability. Extensive digital world experiments are conducted to demonstrate the superiority of the proposed method in the black-box setting. We apply the proposed method in the physical world as well.

To reduce the customer service pressure of small and medium-sized enterprises, we propose an intelligent cloud customer service system, called IntelliTag. Unlike traditional customer service, a cloud service based system has difficulty in collecting user personal information. Therefore, we add a tag recommendation function to quickly capture the user's question intent by clicking on the tags. Specifically, IntelliTag is elaborately designed with the consideration of the following three aspects. First, how to mine high-quality tags is a challenging problem. Second, in the tag recommendation tasks, we have multifarious data types and relations that are used to build a sequential recommendation model. Finally, system implementation and deployment also need to be carefully designed to satisfy online service requirements. In this paper, we show the details of data construction, model designs, system implementation and deployment, and the empirical results compared with several state-of-the-art methods. Nowadays, our IntelliTag has already supported hundreds of thousands of enterprises and millions of users in our industrial production environment.

Recent researches indicate deep learning models are vulnerable to adversarial attacks. Backdoor attack, also called trojan attack, is a variant of adversarial attacks. An malicious attacker can inject backdoor to models in training phase. As a result, the backdoor model performs normally on clean samples and can be triggered by a backdoor pattern to recognize backdoor samples as a wrong target label specified by the attacker. However, the vanilla backdoor attack method causes a measurable difference between clean and backdoor samples in latent space. Several state-of-the-art defense methods utilize this to identify backdoor samples. In this paper, we propose a novel backdoor attack method called SimTrojan, which aims to inject backdoor in models stealthily. Specifically, SimTrojan makes clean and backdoor samples have indistinguishable representations in latent space to evade current defense methods. Experiments demonstrate that SimTrojan achieves a high attack success rate and is undetectable by state-of-the-art defense methods. The study suggests the urgency of building more effective defense methods.

To address the long-standing data sparsity problem in recommender systems (RSs), cross-domain recommendation (CDR) has been proposed to leverage the relatively richer information from a richer domain to improve the recommendation performance in a sparser domain. Although CDR has been extensively studied in recent years, there is a lack of a systematic review of the existing CDR approaches. To fill this gap, in this paper, we provide a comprehensive review of existing CDR approaches, including challenges, research progress, and prospects. Specifically, we first summarize existing CDR approaches into four types, including single-target CDR, single-target multi-domain recommendation (MDR), dual-target CDR, and multi-target CDR. We then present the definitions and challenges of these CDR approaches. Next, we propose a full-view categorization and new taxonomies on these approaches and report their research progress in detail. In the end, we share several promising prospects in CDR.

Logistic Regression (LR) is the most widely used machine learning model in industry for its efficiency, robustness, and interpretability. Due to the problem of data isolation and the requirement of high model performance, many applications in industry call for building a secure and efficient LR model for multiple parties. Most existing work uses either Homomorphic Encryption (HE) or Secret Sharing (SS) to build secure LR. HE based methods can deal with high-dimensional sparse features, but they incur potential security risks. SS based methods have provable security, but they have efficiency issue under high-dimensional sparse features. In this paper, we first present CAESAR, which combines HE and SS to build secure large-scale sparse logistic regression model and achieves both efficiency and security. We then present the distributed implementation of CAESAR for scalability requirement. We have deployed CAESAR in a risk control task and conducted comprehensive experiments. Our experimental results show that CAESAR improves the state-of-the-art model by around 130 times.

Combinatorial Optimization (CO) has been a long-standing challenging research topic featured by its NP-hard nature. Traditionally such problems are approximately solved with heuristic algorithms which are usually fast but may sacrifice the solution quality. Currently, machine learning for combinatorial optimization (MLCO) has become a trending research topic, but most existing MLCO methods treat CO as a single-level optimization by directly learning the end-to-end solutions, which are hard to scale up and mostly limited by the capacity of ML models given the high complexity of CO. In this paper, we propose a hybrid approach to combine the best of the two worlds, in which a bi-level framework is developed with an upper-level learning method to optimize the graph (e.g. add, delete or modify edges in a graph), fused with a lower-level heuristic algorithm solving on the optimized graph. Such a bi-level approach simplifies the learning on the original hard CO and can effectively mitigate the demand for model capacity. The experiments and results on several popular CO problems like Directed Acyclic Graph scheduling, Graph Edit Distance and Hamiltonian Cycle Problem show its effectiveness over manually designed heuristics and single-level learning methods.

Time series forecasting is widely used in business intelligence, e.g., forecast stock market price, sales, and help the analysis of data trend. Most time series of interest are macroscopic time series that are aggregated from microscopic data. However, instead of directly modeling the macroscopic time series, rare literature studied the forecasting of macroscopic time series by leveraging data on the microscopic level. In this paper, we assume that the microscopic time series follow some unknown mixture probabilistic distributions. We theoretically show that as we identify the ground truth latent mixture components, the estimation of time series from each component could be improved because of lower variance, thus benefitting the estimation of macroscopic time series as well. Inspired by the power of Seq2seq and its variants on the modeling of time series data, we propose Mixture of Seq2seq (MixSeq), an end2end mixture model to cluster microscopic time series, where all the components come from a family of Seq2seq models parameterized by different parameters. Extensive experiments on both synthetic and real-world data show the superiority of our approach.

Credit risk prediction is a fundamental problem for most financial institutions.Previous methods mainly adopt users' individual features on a single snapshot.However, users' individual features on financial platforms are usually too sparse to be informative.And previous methods ignore that the features, the behaviours and the credit risk of the users are all dynamic.To resolve the problems, we aim to model the credit risk prediction on dynamic graphs and propose a Temporal-Aware Graph Neural Network (TemGNN) to predict user credit risk.In detail, the model consists of three parts: i) a static model to extract the user's static factors regarding the credit risk.ii) a short-term graph encoder with special graph convolution modules for each snapshot to enrich the user's information through aggregating short-term temporal and structural information.iii) a long-term temporal model based on LSTM with interval-decayed attention to adaptively aggregate the long-term information from the static factors and interval-irregular dynamic snapshots.By combining the three parts together, our model is able to mine both the short-and long-term temporal-structural information.Experimentally, we use the users' authorized lending behaviours as the temporal graphs to do default prediction on Alipay.The results show that our model achieves the best performance among the state-of-the-art methods.

Transfer learning leverages knowledge from a source domain with rich data to a target domain with sparse data. However, the difference between the source and target data distribution weakens the transferability. To bridge this gap, we focus on selecting source instances that are closely related to and have the same distribution as the target domain. In this paper, we propose a novel Adaptive Clustering Transfer Learning (ACTL) method to improve transferability. Specifically, we simultaneously train the instance selector and the transfer learning model. The selector adaptively conducts clustering on the training data and learns the weights for source instances. The weight will activate or inhibit the contribution of the corresponding source instance during transfer learning. Meanwhile, the transfer learning model guides the selector to learn the weight appropriately according to the objective function. To evaluate the effectiveness of our method, we conduct experiments on two different tasks including recommender system and text matching. Experimental results show that our method consistently outperforms competing methods and the selected source instances share a similar data distribution with the target domain.

In the Click-Through Rate (CTR) prediction scenario, user's sequential behaviors are well utilized to capture the user interest in the recent literature. However, despite being extensively studied, these sequential methods still suffer from three limitations. First, existing methods mostly utilize attention on the behavior of users, which is not always suitable for CTR prediction, because users often click on new products that are irrelevant to any historical behaviors. Second, in the real scenario, there are numerous users that have operations a long time ago, but turn relatively inactive in recent times. Thus, it is hard to precisely capture user's current preferences through early behaviors. Third, multiple representations of user's historical behaviors in different feature subspaces are largely ignored. To remedy these issues, we propose a Multi-Interactive Attention Network (MIAN) to comprehensively extract the latent relationship among all kinds of fine-grained features (e.g., gender, age and occupation in user-profile). Specifically, MIAN contains a Multi-Interactive Layer (MIL) that integrates three local interaction modules to capture multiple representations of user preference through sequential behaviors and simultaneously utilize the fine-grained user-specific as well as context information. In addition, we design a Global Interaction Module (GIM) to learn the high-order interactions and balance the different impacts of multiple features. Finally, Offline experiment results from three datasets, together with an Online A/B test in a large-scale recommendation system, demonstrate the effectiveness of our proposed approach.

Recently, Factorization Machine (FM) has become more and more popular for recommendation systems due to its effectiveness in finding informative interactions between features. Usually, the weights for the interactions are learned as a low rank weight matrix, which is formulated as an inner product of two low rank matrices. This low rank matrix can help improve the generalization ability of Factorization Machine. However, to choose the rank properly, it usually needs to run the algorithm for many times using different ranks, which clearly is inefficient for some large-scale datasets. To alleviate this issue, we propose an Adaptive Boosting framework of Factorization Machine (AdaFM), which can adaptively search for proper ranks for different datasets without re-training. Instead of using a fixed rank for FM, the proposed algorithm will gradually increase its rank according to its performance until the performance does not grow. Extensive experiments are conducted to validate the proposed method on multiple large-scale datasets. The experimental results demonstrate that the proposed method can be more effective than the state-of-the-art Factorization Machines.

Network pruning is an important research field aiming at reducing computational costs of neural networks. Conventional approaches follow a fixed paradigm which first trains a large and redundant network, and then determines which units (e.g., channels) are less important and thus can be removed. In this work, we find that pre-training an over-parameterized model is not necessary for obtaining the target pruned structure. In fact, a fully-trained over-parameterized model will reduce the search space for the pruned structure. We empirically show that more diverse pruned structures can be directly pruned from randomly initialized weights, including potential models with better performance. Therefore, we propose a novel network pruning pipeline which allows pruning from scratch with little training overhead. In the experiments for compressing classification models on CIFAR10 and ImageNet datasets, our approach not only greatly reduces the pre-training burden of traditional pruning methods, but also achieves similar or even higher accuracy under the same computation budgets. Our results facilitate the community to rethink the effectiveness of existing techniques used for network pruning.

Bayesian deep learning is recently regarded as an intrinsic way to characterize the weight uncertainty of deep neural networks (DNNs). Stochastic Gradient Langevin Dynamics (SGLD) is an effective method to enable Bayesian deep learning on large-scale datasets. Previous theoretical studies have shown various appealing properties of SGLD, ranging from the convergence properties to the generalization bounds. In this paper, we study the properties of SGLD from a novel perspective of membership privacy protection (i.e., preventing the membership attack). The membership attack, which aims to determine whether a specific sample is used for training a given DNN model, has emerged as a common threat against deep learning algorithms. To this end, we build a theoretical framework to analyze the information leakage (w.r.t. the training dataset) of a model trained using SGLD. Based on this framework, we demonstrate that SGLD can prevent the information leakage of the training dataset to a certain extent. Moreover, our theoretical analysis can be naturally extended to other types of Stochastic Gradient Markov Chain Monte Carlo (SG-MCMC) methods. Empirical results on different datasets and models verify our theoretical findings and suggest that the SGLD algorithm can not only reduce the information leakage but also improve the generalization ability of the DNN models in real-world applications.

With the rapid growth of data volume, data-driven machine learning models have become a necessary part of many industrial applications. Intuitively, the more high-quality data used for training leads to better model performance. However, in reality, data are usually scattered and isolated in different organizations or companies. Such a "data isolation" problem stimulates both academia and industry to explore the collaborative learning paradigm to build better models jointly with multiple data sources. Despite the potential performance gains, this learning paradigm inevitably faces privacy issues, especially for the Fintech domain where data are sensitive by nature. In this paper, we present a privacy-preserving collaborative learning system in Ant Financial, named Nebula. Our system aims to facilitate privacy-preserving collaborative model training for industrial-scale applications. Our system is built upon a ring-allreduce MPI based distributed framework. On top of that, with some optimization strategies and novel sharing scheme, our system is able to scale up to tens of millions of data samples with hundreds of thousands of features and achieve more than 100x speedup compared with the existing state-of-the-art implementations.

Aiming to effectively distinguish loan default in the Mobile Credit Payment Service, industrial efforts mainly attempt to employ conventional classifier with complicated feature engineer for prediction. However, these solutions fail to exploit multiplex relations existed in the financial scenarios and ignore the key intrinsic properties of the loan default detection, i.e., communicability, complementation and induction. To address these issues, we develop a novel attributed multiplex graph based loan default detection approach for effectively integrating multiplex relations in financial scenarios. Considering the complexity of financial scenario, an Attributed Multiplex Graph (AMG) is proposed to jointly model various relations and objects as well as the rich attributes on nodes and edges. We elaborately design relation-specific receptive layers equipped with adaptive breadth function to incorporate important information derived from local structure in each aspect of AMG and stack multiple propagation layer to explore the high-order connectivity information. Furthermore, a relation-specific attention mechanism is adopted to emphasize relevant information during end-to-end training. Extensive experiments conducted on the large-scale real- world dataset verify the effectiveness of the proposed model com- pared with state of arts. Moreover, AMG-DP has also achieved a performance improvement of 9.37% on KS metric in recent months after successful deployment in the Alipay APP.

As one of the core components of customer service bot, User Intent Prediction (UIP) aims at predicting users? intents (usually represented as predefined user questions) before they ask, and has been widely applied in real applications. However, when developing a machine learning system for this problem, two critical issues, i.e., the problem of feature drift and class imbalance, may emerge and seriously deprave the system performance. Moreover, various scenarios may arise due to business demands, making the aforementioned problems much more severe. To address these two problems, we propose an attention-based Deep Multi-instance Sequential Cross Network (aDMSCN) to deal with the UIP task. On the one hand,the UIP task can be subtly formalized as multi-instance learning(MIL) task with an attention-based method proposed to alleviate the influences of feature drift. To the best of our knowledge, this is the first attempt to model the problem from a MIL perspective.On the other hand, a ratio-sensitive loss is also developed in our model, which can mitigate the negative impact of class imbalance. Extensive experiments on both offline real-world datasets and on-line A/B testing show that our proposed framework significantly out performs other state-of-art methods for the UIP task.

Despite the great success of Graph Machine Learning (GML) in a variety of applications, the industry is still seeking a platform which makes performing industrial-purpose GML convenient. In this demo, we present EasyGML, a fully-functional and easy-to-use platform for general AI practitioners to apply out-of-the-box GML models in industrial scenarios. Leveraging the distributed data warehouse as its data infrastructure, EasyGML adopts AGL, an integrated system for industrial-purpose graph learning, as its core GML engine, and develops a model zoo containing various GML models, supporting both node property prediction and link property prediction. It packs different steps of GML workflow into different components, and provides a user-friendly web-based GUI for users to build their GML workflows simply by connecting several components together, without any coding.

Nowadays, privacy preserving machine learning has been drawing much attention in both industry and academy. Meanwhile, recommender systems have been extensively adopted by many commercial platforms (e.g. Amazon) and they are mainly built based on user-item interactions. Besides, social platforms (e.g. Facebook) have rich resources of user social information. It is well known that social information, which is rich on social platforms such as Facebook, are useful to build intelligent recommender systems. It is anticipated to combine the social information with the user-item ratings to improve the overall recommendation performance. Most existing recommendation models are built based on the assumptions that the social information are available. However, different platforms are usually reluctant to (or can not) share their data due to certain concerns. In this paper, we first propose a SEcure SOcial RECommendation (SeSoRec) framework which is able to (1) collaboratively mine knowledge from social platform to improve the recommendation performance of the rating platform, and (2) securely keep the raw data of both platforms. We then propose a Secret Sharing based Matrix Multiplication (SSMM) protocol to optimize SeSoRec and prove its correctness and security theoretically. By applying minibatch gradient descent, SeSoRec has linear time complexities in terms of both computation and communication. The comprehensive experimental results on three real-world datasets demonstrate the effectiveness of our proposed SeSoRec and SSMM.

Today text classification models have been widely used. However, these classifiers are found to be easily fooled by adversarial examples. Fortunately, standard attacking methods generate adversarial texts in a pair-wise way, that is, an adversarial text can only be created from a real-world text by replacing a few words. In many applications, these texts are limited in numbers, therefore their corresponding adversarial examples are often not diverse enough and sometimes hard to read, thus can be easily detected by humans and cannot create chaos at a large scale. In this paper, we propose an end to end solution to efficiently generate adversarial texts from scratch using generative models, which are not restricted to perturbing the given texts. We call it unrestricted adversarial text generation. Specifically, we train a conditional variational autoencoder (VAE) with an additional adversarial loss to guide the generation of adversarial examples. Moreover, to improve the validity of adversarial texts, we utilize discrimators and the training framework of generative adversarial networks (GANs) to make adversarial texts consistent with real data. Experimental results on sentiment analysis demonstrate the scalability and efficiency of our method. It can attack text classification models with a higher success rate than existing methods, and provide acceptable quality for humans in the meantime.

Machine learning techniques have been widely applied in Internet companies for various tasks, acting as an essential driving force, and feature engineering has been generally recognized as a crucial tache when constructing machine learning systems. Recently, a growing effort has been made to the development of automatic feature engineering methods, so that the substantial and tedious manual effort can be liberated. However, for industrial tasks, the efficiency and scalability of these methods are still far from satisfactory. In this paper, we proposed a staged method named SAFE (Scalable Automatic Feature Engineering), which can provide excellent efficiency and scalability, along with requisite interpretability and promising performance. Extensive experiments are conducted and the results show that the proposed method can provide prominent efficiency and competitive effectiveness when comparing with other methods. What's more, the adequate scalability of the proposed method ensures it to be deployed in large scale industrial tasks.

Small and Medium-sized Enterprises (SMEs) are playing a vital role in the modern economy. Recent years, financial risk analysis for SMEs attracts lots of attentions from financial institutions. However, the financial risk analysis for SMEs usually suffers data deficiency problem, especially for the mobile financial institutions which seldom collect credit-related data directly from SMEs. Fortunately, although credit-related information of SMEs is hard to be acquired sufficiently, the interactive relationships between SMEs, which may contain valuable information of financial risk, is usually available for the mobile financial institutions. Finding out credit-related relationship of SME from massive interactions helps comprehensively model the SMEs thus improve the performance of financial risk analysis. In this paper, tackling the data deficiency problem of financial risk analysis for SMEs, we propose an innovative financial risk analysis framework with graph-based supply chain mining. Specifically, to capture the credit-related topology structural and temporal variation information of SMEs, we design and employ a novel spatial-temporal aware graph neural network, to mine supply chain relationship on a SME graph, and then analysis the credit risk based on the mined supply chain graph. Experimental results on real-world financial datasets prove the effectiveness of our proposal for financial risk analysis for SMEs.

Recently, in order to take a preemptive opportunity in the mobile economy, the Internet companies conduct thousands of marketing campaigns every day, to promote their mobile products and services. In the mobile marketing scenario, one of the fundamental issues is the audience expansion task for marketing campaigns. Given a set of seed users, audience expansion aims to seek more users (audiences), who are similar to the seeds and will finish the business goal of the targeted campaign (ie convert). However, the problem is challenging in three aspects. First, a company will run hundreds of campaigns to serve massive users every day. The requirements of scalability and timeliness make training model for each campaign extremely resource-consuming thus impractical. Therefore, we proposed to solve the problem in a two-stage manner, in which the offline stage employs heavyweight user representation learning and the online stage performs embedding-based lightweight audience expansion. Second, conventional two-stage audience expansion systems neglect the high-order user-campaign interactions and usually generate entangled user embeddings, thus fail to achieve high-quality user representation. Third, the seeds, which are usually provided by experts or collected from users' feedbacks, could be noisy and cannot cover the entire actual audiences, thus introduce coverage bias. Unfortunately, to our best knowledge, none of the related literatures tackle this crucial issue of audience expansion.

Several sampling algorithms with variance reduction have been proposed for accelerating the training of Graph Convolution Networks (GCNs). However, due to the intractable computation of optimal sampling distribution, these sampling algorithms are suboptimal for GCNs and are not applicable to more general graph neural networks (GNNs) where the message aggregator contains learned weights rather than fixed weights, such as Graph Attention Networks (GAT). The fundamental reason is that the embeddings of the neighbors or learned weights involved in the optimal sampling distribution are \emph{changing} during the training and \emph{not known a priori}, but only \emph{partially observed} when sampled, thus making the derivation of an optimal variance reduced samplers non-trivial. In this paper, we formulate the optimization of the sampling variance as an adversary bandit problem, where the rewards are related to the node embeddings and learned weights, and can vary constantly. Thus a good sampler needs to acquire variance information about more neighbors (exploration) while at the same time optimizing the immediate sampling variance (exploit). We theoretically show that our algorithm asymptotically approaches the optimal variance within a factor of 3. We show the efficiency and effectiveness of our approach on multiple datasets.

Machine learning over graphs has been emerging as powerful learning tools for graph data. However, it is challenging for industrial communities to leverage the techniques, such as graph neural networks (GNNs), and solve real-world problems at scale because of inherent data dependency in the graphs. As such, we cannot simply train a GNN with classic learning systems, for instance, parameter server that assumes data parallelism. Existing systems store the graph data in-memory for fast accesses either in a single machine or graph stores from remote. The major drawbacks are three-fold. First, they cannot scale because of the limitations on the volume of the memories, or the bandwidth between graph stores and workers. Second, they require extra development of graph stores without well exploiting mature infrastructures such as MapReduce that guarantee good system properties. Third, they focus on training but ignore optimizing the performance of inference over graphs, thus makes them an unintegrated system. In this paper, we design AGL, a scalable and integrated system, with fully-functional training and inference for GNNs. Our system design follows the message passing scheme underlying the computations of GNNs. We design to generate the K -hop neighborhood, an information-complete subgraph for each node, as well as do the inference simply by merging values from in-edge neighbors and propagating values to out-edge neighbors via MapReduce. In addition, the K -hop neighborhood contains information-complete subgraphs for each node, thus we simply do the training on parameter servers due to data independence. Our system AGL, implemented on mature infrastructures, can finish the training of a 2-layer GNN on a graph with billions of nodes and hundred billions of edges in 14 hours, and complete the inference in 1.2 hours.

Point-of-Interest (POI) recommendation has been extensively studied and successfully applied in industry recently. However, most existing approaches build centralized models on the basis of collecting users’ data. Both private data and models are held by the recommender, which causes serious privacy concerns. In this article, we propose a novel Privacy preserving POI Recommendation (PriRec) framework. First, to protect data privacy, users’ private data (features and actions) are kept on their own side, e.g., Cellphone or Pad. Meanwhile, the public data that need to be accessed by all the users are kept by the recommender to reduce the storage costs of users’ devices. Those public data include: (1) static data only related to the status of POI, such as POI categories, and (2) dynamic data dependent on user-POI actions such as visited counts. The dynamic data could be sensitive, and we develop local differential privacy techniques to release such data to the public with privacy guarantees. Second, PriRec follows the representations of Factorization Machine (FM) that consists of a linear model and the feature interaction model. To protect the model privacy, the linear models are saved on the users’ side, and we propose a secure decentralized gradient descent protocol for users to learn it collaboratively. The feature interaction model is kept by the recommender since there is no privacy risk, and we adopt a secure aggregation strategy in a federated learning paradigm to learn it. To this end, PriRec keeps users’ private raw data and models in users’ own hands, and protects user privacy to a large extent. We apply PriRec in real-world datasets, and comprehensive experiments demonstrate that, compared with FM, PriRec achieves comparable or even better recommendation accuracy.

As one of the major frauds in financial services, cash-out fraud is that users pursue cash gains with illegal or insincere means. Conventional solutions for the cash-out user detection are to perform subtle feature engineering for each user and then apply a classifier, such as GDBT and Neural Network. However, users in financial services have rich interaction relations, which are seldom fully exploited by conventional solutions. In this paper, with the real datasets in Ant Credit Pay of Ant Financial Services Group, we first study the cashout user detection problem and propose a novel hierarchical attention mechanism based cash-out user detection model, called HACUD. Specifically, we model different types of objects and their rich attributes and interaction relations in the scenario of credit payment service with an Attributed Heterogeneous Information Network (AHIN). The HACUD model enhances feature representation of objects through meta-path based neighbors exploiting different aspects of structure information in AHIN. Furthermore, a hierarchical attention mechanism is elaborately designed to model user’s preferences towards attributes and meta-paths. Experimental results on two real datasets show that the HACUD outperforms the state-of-the-art methods.

We present, GeniePath, a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In GeniePath, we propose an adaptive path layer consists of two complementary functions designed for breadth and depth exploration respectively, where the former learns the importance of different sized neighborhoods, while the latter extracts and filters signals aggregated from neighbors of different hops away. Our method works in both transductive and inductive settings, and extensive experiments compared with competitive methods show that our approaches yield state-of-the-art results on large graphs.

The insurance industry has been creating innovative products around the emerging online shopping activities. Such ecommerce insurance is designed to protect buyers from potential risks such as impulse purchases and counterfeits. Fraudulent claims towards online insurance typically involve multiple parties such as buyers, sellers, and express companies, and they could lead to heavy financial losses. In order to uncover the relations behind organized fraudsters and detect fraudulent claims, we developed a large-scale insurance fraud detection system, i.e., InfDetect, which provides interfaces for commonly used graphs, standard data processing procedures, and a uniform graph learning platform. InfDetect is able to process big graphs containing up to 100 millions of nodes and billions of edges.In this paper, we investigate different graphs to facilitate fraudster mining, such as a device-sharing graph, a transaction graph, a friendship graph, and a buyer-seller graph. These graphs are fed to a uniform graph learning platform containing supervised and unsupervised graph learning algorithms. Cases on widely applied e-commerce insurance are described to demonstrate the usage and capability of our system. InfDetect has successfully detected thousands of fraudulent claims and saved over tens of thousands of dollars daily.

Secure online transaction is an essential task for e-commerce platforms. Alipay, one of the world’s leading cashless payment platform, provides the payment service to both merchants and individual customers. The fraud detection models are built to protect the customers, but stronger demands are raised by the new scenes, which are lacking in training data and labels. The proposed model makes a difference by utilizing the data under similar old scenes and the data under a new scene is treated as the target domain to be promoted. Inspired by this real case in Alipay, we view the problem as a transfer learning problem and design a set of revise strategies to transfer the source domain models to the target domain under the framework of gradient boosting tree models. This work provides an option for the cold-start and data-sharing problems.

Link prediction is widely used in a variety of industrial applications, such as merchant recommendation, fraudulent transaction detection, and so on. However, it's a great challenge to train and deploy a link prediction model on industrial-scale graphs with billions of nodes and edges. In this work, we present a scalable and distributed framework for semi-supervised link prediction problem (named DSSLP), which is able to handle industrial-scale graphs. Instead of training model on the whole graph, DSSLP is proposed to train on the k-hops neighborhood of nodes in a mini-batch setting, which helps reduce the scale of the input graph and distribute the training procedure. In order to generate negative examples effectively, DSSLP contains a distributed batched runtime sampling module. It implements uniform and dynamic sampling approaches, and is able to adaptively construct positive and negative examples to guide the training process. Moreover, DSSLP proposes a model-split strategy to accelerate the speed of inference process of the link prediction task. Experimental results demonstrate that the effectiveness and efficiency of DSSLP in serval public datasets as well as real-world datasets of industrial-scale graphs.

Mobile payment such as Alipay has been widely used in our daily lives. To further promote the mobile payment activities, it is important to run marketing campaigns under a limited budget by providing incentives such as coupons, commissions to merchants. As a result, incentive optimization is the key to maximizing the commercial objective of the marketing campaign. With the analyses of online experiments, we found that the transaction network can subtly describe the similarity of merchants' responses to different incentives, which is of great use in the incentive optimization problem. In this paper, we present a graph representation learning method atop of transaction networks for merchant incentive optimization in mobile payment marketing. With limited samples collected from online experiments, our end-to-end method first learns merchant representations based on an attributed transaction networks, then effectively models the correlations between the commercial objectives each merchant may achieve and the incentives under varying treatments. Thus we are able to model the sensitivity to incentive for each merchant, and spend the most budgets on those merchants that show strong sensitivities in the marketing campaign. Extensive offline and online experimental results at Alipay demonstrate the effectiveness of our proposed approach.

With the rapid growth of financial services, fraud detection has been a very important problem to guarantee a healthy environment for both users and providers. Conventional solutions for fraud detection mainly use some rule-based methods or distract some features manually to perform prediction. However, in financial services, users have rich interactions and they themselves always show multifaceted information. These data form a large multiview network, which is not fully exploited by conventional methods. Additionally, among the network, only very few of the users are labelled, which also poses a great challenge for only utilizing labeled data to achieve a satisfied performance on fraud detection. To address the problem, we expand the labeled data through their social relations to get the unlabeled data and propose a semi-supervised attentive graph neural network, named SemiGNN to utilize the multi-view labeled and unlabeled data for fraud detection. Moreover, we propose a hierarchical attention mechanism to better correlate different neighbors and different views. Simultaneously, the attention mechanism can make the model interpretable and tell what are the important factors for the fraud and why the users are predicted as fraud. Experimentally, we conduct the prediction task on the users of Alipay, one of the largest third-party online and offline cashless payment platform serving more than 4 hundreds of million users in China. By utilizing the social relations and the user attributes, our method can achieve a better accuracy compared with the state-of-the-art methods on two tasks. Moreover, the interpretable results also give interesting intuitions regarding the tasks.

We create an intention mining system, named AntProphet, for Alipay's intelligent customer service bot, to alleviate the burden of customer service. Whenever users have any questions, AntProphet is the first stop to help users to answer their questions. Our system gathers users' profile and their historical behavioral trajectories, together with contextual information to predict users' intention, i.e., the potential questions that users want to resolve. AntProphet takes care of more than 90% of the customer service demands in the Alipay APP and resolves most of the users' problems on the spot, thus significantly reduces the burden of manpower. With the help of it, the overall satisfaction rate of our customer service bot exceeds 85%.

In this paper, we aim to understand the generalization properties of generative adversarial networks (GANs) from a new perspective of privacy protection. Theoretically, we prove that a differentially private learning algorithm used for training the GAN does not overfit to a certain degree, i.e., the generalization gap can be bounded. Moreover, some recent works, such as the Bayesian GAN, can be re-interpreted based on our theoretical insight from privacy protection. Quantitatively, to evaluate the information leakage of well-trained GAN models, we perform various membership attacks on these models. The results show that previous Lipschitz regularization techniques are effective in not only reducing the generalization gap but also alleviating the information leakage of the training dataset.

A customer service bot is now a necessary component of an e-commerce platform. As a core module of the customer service bot, user intent prediction can help predict user questions before they ask. A typical solution is to find top candidate questions that a user will be interested in. Such solution ignores the inter-relationship between questions and often aims to maximize the immediate reward such as clicks, which may not be ideal in practice. Hence, we propose to view the problem as a sequential decision making process to better capture the long-term effects of each recommendation in the list. Intuitively, we formulate the problem as a Markov decision process and consider using reinforcement learning for the problem. With this approach, questions presented to users are both relevant and diverse. Experiments on offline real-world dataset and online system demonstrate the effectiveness of our proposed approach.

Fraudulent claim detection is one of the greatest challenges the insurance industry faces. Alibaba's return-freight insurance, providing return-shipping postage compensations over product return on the e-commerce platform, receives thousands of potentially fraudulent claims everyday. Such deliberate abuse of the insurance policy could lead to heavy financial losses. In order to detect and prevent fraudulent insurance claims, we developed a novel data-driven procedure to identify groups of organized fraudsters, one of the major contributions to financial losses, by learning network information. In this paper, we introduce a device-sharing network among claimants, followed by developing an automated solution for fraud detection based on graph learning algorithms, to separate fraudsters from regular customers and uncover groups of organized fraudsters. This solution applied at Alibaba achieves more than 80% precision while covering 44% more suspicious accounts compared with a previously deployed rule-based classifier after human expert investigations. Our approach can easily and effectively generalizes to other types of insurance.

Consumers today face too many reviews to read when shopping online. Presenting the most helpful reviews, instead of all, to them will greatly ease purchase decision making. Most of the existing studies on review helpfulness prediction focused on domains with rich labels, not suitable for domains with insufficient labels. In response, we explore a multi-domain approach that learns domain relationships to help the task by transferring knowledge from data-rich domains to data-deficient domains. To better model domain differences, our approach gates multi-granularity embeddings in a Neural Network (NN) based transfer learning framework to reflect the domain-variant importance of words. Extensive experiments empirically demonstrate that our model outperforms the state-of-the-art baselines and NN-based methods without gating on this task. Our approach facilitates more effective knowledge transfer between domains, especially when the target domain dataset is small. Meanwhile, the domain relationship and domain-specific embedding gating are insightful and interpretable.

With the explosive growth of e-commerce and the booming of e-payment, detecting online transaction fraud in real time has become increasingly important to Fintech business. To tackle this problem, we introduce the TitAnt, a transaction fraud detection system deployed in Ant Financial, one of the largest Fintech companies in the world. The system is able to predict online real-time transaction fraud in mere milliseconds. We present the problem definition, feature extraction, detection methods, implementation and deployment of the system, as well as empirical effectiveness. Extensive experiments have been conducted on large real-world transaction data to show the effectiveness and the efficiency of the proposed system.

Internet companies are facing the need for handling large-scale machine learning applications on a daily basis and distributed implementation of machine learning algorithms which can handle extra-large-scale tasks with great performance is widely needed. Deep forest is a recently proposed deep learning framework which uses tree ensembles as its building blocks and it has achieved highly competitive results on various domains of tasks. However, it has not been tested on extremely large-scale tasks. In this work, based on our parameter server system, we developed the distributed version of deep forest. To meet the need for real-world tasks, many improvements are introduced to the original deep forest model, including MART (Multiple Additive Regression Tree) as base learners for efficiency and effectiveness consideration, the cost-based method for handling prevalent class-imbalanced data, MART based feature selection for high dimension data, and different evaluation metrics for automatically determining the cascade level. We tested the deep forest model on an extra-large-scale task, i.e., automatic detection of cash-out fraud, with more than 100 million training samples. Experimental results showed that the deep forest model has the best performance according to the evaluation metrics from different perspectives even with very little effort for parameter tuning. This model can block fraud transactions in a large amount of money each day. Even compared with the best-deployed model, the deep forest model can additionally bring a significant decrease in economic loss each day.

We propose cw2vec, a novel method for learning Chinese word embeddings. It is based on our observation that exploiting stroke-level information is crucial for improving the learning of Chinese word embeddings. Specifically, we design a minimalist approach to exploit such features, by using stroke n-grams, which capture semantic and morphological level information of Chinese words. Through qualitative analysis, we demonstrate that our model is able to extract semantic information that cannot be captured by existing methods. Empirical results on the word similarity, word analogy, text classification and named entity recognition tasks show that the proposed approach consistently outperforms state-of-the-art approaches such as word-based word2vec and GloVe, character-based CWE, component-based JWE and pixel-based GWE.

Points of interest (POI) recommendation has been drawn much attention recently due to the increasing popularity of location-based networks, e.g., Foursquare and Yelp. Among the existing approaches to POI recommendation, Matrix Factorization (MF) based techniques have proven to be effective. However, existing MF approaches suffer from two major problems: (1) Expensive computations and storages due to the centralized model training mechanism: the centralized learners have to maintain the whole user-item rating matrix, and potentially huge low rank matrices. (2) Privacy issues: the users' preferences are at risk of leaking to malicious attackers via the centralized learner. To solve these, we present a Decentralized MF (DMF) framework for POI recommendation. Specifically, instead of maintaining all the low rank matrices and sensitive rating data for training, we propose a random walk based decentralized training technique to train MF models on each user's end, e.g., cell phone and Pad. By doing so, the ratings of each user are still kept on one's own hand, and moreover, decentralized learning can be taken as distributed learning with multi-learners (users), and thus alleviates the computation and storage issue. Experimental results on two real-world datasets demonstrate that, comparing with the classic and state-of-the-art latent factor models, DMF significantly improvements the recommendation performance in terms of precision and recall.

Ant Credit Pay is a consumer credit service in Ant Financial Service Group. Similar to credit card, loan default is one of the major risks of this credit product. Hence, effective algorithm for default prediction is the key to losses reduction and profits increment for the company. However, the challenges facing in our scenario are different from those in conventional credit card service. The first one is scalability. The huge volume of users and their behaviors in Ant Financial requires the ability to process industrial-scale data and perform model training efficiently. The second challenges is the cold-start problem. Different from the manual review for credit card application in conventional banks, the credit limit of Ant Credit Pay is automatically offered to users based on the knowledge learned from big data. However, default prediction for new users is suffered from lack of enough credit behaviors. It requires that the proposal should leverage other new data source to alleviate the cold-start problem. Considering the above challenges and the special scenario in Ant Financial, we try to incorporate default prediction with network information to alleviate the cold-start problem. In this paper, we propose an industrial-scale distributed network representation framework, termed NetDP, for default prediction in Ant Credit Pay. The proposal explores network information generated by various interaction between users, and blends unsupervised and supervised network representation in a unified framework for default prediction problem. Moreover, we present a parameter-server-based distributed implement of our proposal to handle the scalability challenge. Experimental results demonstrate the effectiveness of our proposal, especially in cold-start problem, as well as the efficiency for industrial-scale dataset.

We present, GEM, the first heterogeneous graph neural network approach for detecting malicious accounts at Alipay, one of the world's leading mobile cashless payment platform. Our approach, inspired from a connected subgraph approach, adaptively learns discriminative embeddings from heterogeneous account-device graphs based on two fundamental weaknesses of attackers, i.e. device aggregation and activity aggregation. For the heterogeneous graph consists of various types of nodes, we propose an attention mechanism to learn the importance of different types of nodes, while using the sum operator for modeling the aggregation patterns of nodes in each type. Experiments show that our approaches consistently perform promising results compared with competitive methods over time.

Collaborative filtering, especially latent factor model, has been popularly used in personalized recommendation. Latent factor model aims to learn user and item latent factors from user-item historic behaviors. To apply it into real big data scenarios, efficiency becomes the first concern, including offline model training efficiency and online recommendation efficiency. In this paper, we propose a D istributed C ollaborative H ashing ( DCH ) model which can significantly improve both efficiencies. Specifically, we first propose a distributed learning framework, following the state-of-the-art parameter server paradigm, to learn the offline collaborative model. Our model can be learnt efficiently by distributedly computing subgradients in minibatches on workers and updating model parameters on servers asynchronously. We then adopt hashing technique to speedup the online recommendation procedure. Recommendation can be quickly made through exploiting lookup hash tables. We conduct thorough experiments on two real large-scale datasets. The experimental results demonstrate that, comparing with the classic and state-of-the-art (distributed) latent factor models, DCH has comparable performance in terms of recommendation accuracy but has both fast convergence speed in offline model training procedure and realtime efficiency in online recommendation procedure. Furthermore, the encouraging performance of DCH is also shown for several real-world applications in Ant Financial.

Nowadays, the number of mobile applications (Apps) is rapidly increasing due to different kinds of user needs. Existing App recommendation models can not scale to practical large data scenarios. In this paper, we present a large scale App recommendation framework, which has been applied into real application in Ant Financial for recommending Apps for hundreds of millions of users. We first present feature engineering, including a novel way to high-level features, and then use a linear ranking model to make recommendation. We also report the encouraging results through the experiments in real large scale dataset.

In this work, we propose to actively detect implicit fraudulent transactions. A novel machine learning method is introduced to distinguish anomalous electronic transactions based on the historical records. The transferor will be alerted during the on-going payment when the fraud probability is recognized as large enough. Compared with elaborative rule-based approaches, our model is much more effective in fraud detection.

Nowadays, online payment systems play more and more important roles in people's daily lives. A key component of these systems is to detect and prevent fraud transactions. In industrial practice, such a task is separated into two phases: 1) mining evidential features to describe users, 2) building an effective model based on these features. Generally speaking, the most popular fraud transaction detection systems use elaborately designed features to build tree based models, sometimes a subsequent linear model is added to improve the behaviour. However, the designed features usually contains only static features, while dynamic features are not considered. In addition, the subsequent model can only learn a linear combination, which may always be unsatisfactory. To address these issues, we present a systematic method, which extracts not only users' static features but also dynamic features based on their recent behaviors. Moreover, N-GRAM model is employed to handle the dynamic features so that time series information is addressed. Based on the extracted features, a tree based model is applied and the outputs of it are regarded as new generated feature representations, which will be further inputted into a Deep Neural Network (DNN) to learn the complex relationships and form the final classification model. Extensive experiments show that our proposed model (with both static and dynamic features) significantly outperforms the existing methods.

We present a neural network based graph embedding method for detecting malicious accounts at Alipay, one of the world's leading mobile payment platform. Our method adaptively learns discriminative embeddings from an account-device graph based on two fundamental weaknesses of attackers, i.e. device aggregation and activity aggregation. Experiments show that our method achieves outstanding precision-recall curve compared with existing methods.

This paper describes a PU learning (Positive and Unlabeled learning) based system for potential URL attack detection. Previous machine learning based solutions for this task mainly formalize it as a supervised learning problem. However, in some scenarios, the data obtained always contains only a handful of known attack URLs, along with a large number of unlabeled instances, making the supervised learning paradigms infeasible. In this work, we formalize this setting as a PU learning problem, and solve it by combining two different strategies (two-stage strategy and cost-sensitive strategy). Experimental results show that the developed system can effectively find potential URL attacks. This system can either be deployed as an assistance for existing system or be employed to help cyber-security engineers to effectively discover potential attack mode so that they can improve the existing system with significantly less efforts.

Factorization Machines (FMs) are a widely used method for efficiently using high-order feature interactions in classification and regression tasks. Unfortunately, despite increasing interests in FMs, existing work only considers high order information of the input features which limits their capacities in non-linear problems and fails to capture the underlying structures of more complex data. In this work, we present a novel Locally Linear Factorization Machines (LLFM) which overcomes this limitation by exploring local coding technique. Unlike existing local coding classifiers that involve a phase of unsupervised anchor point learning and predefined local coding scheme which is suboptimal as the class label information is not exploited in discovering the encoding and thus can result in a suboptimal encoding for prediction, we formulate a joint optimization over the anchor points, local coding coordinates and FMs variables to minimize classification or regression risk. Empirically, we demonstrate that our approach achieves much better predictive accuracy than other competitive methods which employ LLFM with unsupervised anchor point learning and predefined local coding scheme.

In recent years, due to the emergence of Big Data (terabytes or petabytes) and Big Model (tens of billions of parameters), there has been an ever-increasing need of parallelizing machine learning (ML) algorithms in both academia and industry. Although there are some existing distributed computing systems, such as Hadoop and Spark, for parallelizing ML algorithms, they only provide synchronous and coarse-grained operators (e.g., Map, Reduce, and Join, etc.), which may hinder developers from implementing more efficient algorithms. This motivated us to design a universal distributed platform termed KunPeng, that combines both distributed systems and parallel optimization algorithms to deal with the complexities that arise from large-scale ML. Specifically, KunPeng not only encapsulates the characteristics of data/model parallelism, load balancing, model sync-up, sparse representation, industrial fault-tolerance, etc., but also provides easy-to-use interface to empower users to focus on the core ML logics. Empirical results on terabytes of real datasets with billions of samples and features demonstrate that, such a design brings compelling performance improvements on ML programs ranging from Follow-the-Regularized-Leader Proximal algorithm to Sparse Logistic Regression and Multiple Additive Regression Trees. Furthermore, KunPeng's encouraging performance is also shown for several real-world applications including the Alibaba's Double 11 Online Shopping Festival and Ant Financial's transaction risk estimation.