Vision-Language Models (VLMs) demand substantial computational resources during inference, largely due to the extensive visual input tokens for representing visual information. Previous studies have noted that visual tokens tend to receive less attention than text tokens, suggesting their lower importance during inference and potential for pruning. However, their methods encounter several challenges: reliance on greedy heuristic criteria for token importance and incompatibility with FlashAttention and KV cache. To address these issues, we introduce TopV, a compatible TOken Pruning with inference Time Optimization for fast and low-memory VLM, achieving efficient pruning without additional training or fine-tuning. Instead of relying on attention scores, we formulate token pruning as an optimization problem, accurately identifying important visual tokens while remaining compatible with FlashAttention. Additionally, since we only perform this pruning once during the prefilling stage, it effectively reduces KV cache size. Our optimization framework incorporates a visual-aware cost function considering factors such as Feature Similarity, Relative Spatial Distance, and Absolute Central Distance, to measure the importance of each source visual token, enabling effective pruning of low-importance tokens. Extensive experiments demonstrate that our method outperforms previous token pruning methods, validating the effectiveness and efficiency of our approach.

Densely structured pruning methods — which generate pruned models in a fully dense format, allowing immediate compression benefits without additional demands — are evolving due to their practical significance. Traditional techniques in this domain mainly revolve around coarser granularities, such as filter pruning, thereby limiting performance due to restricted pruning freedom. Recent advancements in Grouped Kernel Pruning (GKP) have enabled the utilization of finer granularities while maintaining a densely structured format. We observe that existing GKP methods often introduce dynamic operations to different aspects of their procedures at the cost of adding complications and/or imposing limitations (e.g., requiring an expensive mixture of clustering schemes), or contain dynamic pruning rates and sizes among groups that result in a reliance on custom architecture support for its pruned models. In this work, we argue that the best practice to introduce these dynamic operations to GKP is to make Conv2d(groups) (a.k.a. group count) flexible under an integral optimization, leveraging its ideal alignment with the infrastructure support of Grouped Convolution. Pursuing such a direction, we present a one-shot, post-train, data-agnostic GKP method that is more performant, adaptive, and efficient than its predecessors while simultaneously being user-friendly, with little-to-no hyper-parameter tuning or handcrafting of criteria required.

Large language models (LLMs) can give out erroneous answers to factually rooted questions either as a result of undesired training outcomes or simply because the world has moved on after a certain knowledge cutoff date. Under such scenarios, *knowledge editing* often comes to the rescue by delivering efficient patches for such erroneous answers without significantly altering the rest, where many editing methods have seen reasonable success when the editing targets are simple and direct (e.g., *``what club does Lionel Messi currently play for?''*). However, knowledge fragments like this are often deeply intertwined in the real world, making effectively propagating the editing effect to non-directly related questions a practical challenge (to entertain an extreme example: [*"What car did the wife of the owner of the club that Messi currently plays for used to get to school in the 80s?"*](youtube.com/watch?v=DbwiHC1Fu-E\&t=132s)). Prior arts have coined this task as *multi-hop knowledge editing* with the most popular dataset being MQuAKE, serving as the sole evaluation benchmark for many later proposed editing methods due to the expensive nature of constructing knowledge editing datasets at scale. In this work, we reveal that **up to 33\% or 76\% of \mquake{}'s questions and ground truth labels are, in fact, corrupted in various fashions due to some unintentional clerical or procedural oversights**. Our work provides a detailed audit of MQuAKE's error pattern and a comprehensive fix without sacrificing its dataset capacity. Additionally, we benchmarked almost all proposed MQuAKE-evaluated editing methods on our post-fix dataset, **MQuAKE-Remastered**. We observe that many methods try to overfit the original MQuAKE by exploiting some dataset idiosyncrasies of MQuAKE. We provide a guideline on how to approach such datasets faithfully and show that a simple, minimally invasive approach — **GWalk** — can offer beyond SOTA editing performance without such exploitation. The MQuAKE-Remastered datasets and utilities are available at [huggingface.co/datasets/henryzhongsc/MQuAKE-Remastered](https://huggingface.co/datasets/henryzhongsc/MQuAKE-Remastered) and [github.com/henryzhongsc/MQuAKE-Remastered](https://github.com/henryzhongsc/MQuAKE-Remastered), respectively.

Large language models (LLMs) have demonstrated impressive performance on several tasks and are increasingly deployed in real-world applications. However, especially in high-stakes settings, it becomes vital to know when the output of an LLM may be unreliable. Depending on whether an answer is trustworthy, a system can then choose to route the question to another expert, or otherwise fall back on a safe default behavior. In this work, we study the extent to which LLMs can reliably indicate confidence in their answers, and how this notion of confidence can translate into downstream accuracy gains. We propose Self-Reflection with Error-based Feedback (Self-REF), a lightweight training strategy to teach LLMs to express confidence in whether their answers are correct in a reliable manner. Self-REF introduces confidence tokens into the LLM, from which a confidence score can be extracted. Compared to conventional approaches such as verbalizing confidence and examining token probabilities, we demonstrate empirically that confidence tokens show significant improvements in downstream routing and rejection learning tasks.

Large Language Models (LLMs) are now integral across various domains and have demonstrated impressive performance. Progress, however, rests on the premise that benchmark scores are both accurate and reproducible. We demonstrate that the reproducibility of LLM performance is fragile: changing system configuration, such as evaluation batch size, GPU count, and GPU version, can introduce significant differences in the generated responses. This issue is especially pronounced in reasoning models, where minor rounding differences in early tokens can cascade into divergent chains of thought, ultimately affecting accuracy. For instance, under bfloat16 precision with greedy decoding, a reasoning model like DeepSeek-R1-Distill-Qwen-7B can exhibit up to 9\% variation in accuracy and 9,000 tokens difference in response length due to differences in GPU count, type, and evaluation batch size. We trace the root cause of this variability to the non-associative nature of floating-point arithmetic under limited numerical precision. This work presents the first systematic investigation into how numerical precision affects reproducibility in LLM inference. Through carefully controlled experiments across various hardware, software, and precision settings, we quantify when and how model outputs diverge. Our analysis reveals that floating-point precision—while critical for reproducibility—is often neglected in evaluation practices. Inspired by this, we develop a lightweight inference pipeline, dubbed LayerCast, that stores weights in 16-bit precision but performs all computations in FP32, balancing memory efficiency with numerical stability. Code is available at https://github.com/nanomaoli/llm_reproducibility.

Large-scale AI models, such as Large Language Models (LLMs) and Diffusion Models (DMs), have grown rapidly in size, creating significant challenges for efficient deployment on resource-constrained hardware. In this paper, we introduce Dynamic-Length Float (DFloat11), a lossless compression framework that reduces LLM and DM size by 30\% while preserving outputs that are bit-for-bit identical to the original model. DFloat11 is motivated by the low entropy in the BFloat16 weight representation of LLMs, which reveals significant inefficiency in the existing storage format. By applying entropy coding, DFloat11 assigns dynamic-length encodings to weights based on frequency, achieving near information-optimal compression without any loss of precision. To facilitate efficient inference with dynamic-length encodings, we develop a custom GPU kernel for fast online decompression. Our design incorporates the following: (i) compact, hierarchical lookup tables (LUTs) that fit within GPU SRAM for efficient decoding, (ii) a two-phase GPU kernel for coordinating thread read/write positions using lightweight auxiliary variables, and (iii) transformer-block-level decompression to minimize latency. Experiments on Llama 3.3, Qwen 3, Mistral 3, FLUX.1, and others validate our hypothesis that DFloat11 achieves around 30\% model size reduction while preserving bit-for-bit identical outputs. Compared to a potential alternative of offloading parts of an uncompressed model to the CPU to meet memory constraints, DFloat11 achieves 2.3--46.2$\times$ higher throughput in token generation. With a fixed GPU memory budget, DFloat11 enables 5.7--14.9$\times$ longer generation lengths than uncompressed models. Notably, our method enables lossless inference of Llama 3.1 405B, an 810GB model, on a single node equipped with 8$\times$80GB GPUs.

OBJECTIVES: Primary graft dysfunction (PGD) is an essential outcome after the heart transplant, which causes severe complications and symptoms for recipients. The in advance prediction of PGD can help the transplant physician better manage the risks of PGD occurrence for patients. Domain experts have identified some important risk factors leading to PGD. However, a widely accepted PGD prediction method is lacking from a computational perspective. In this work, we focus on the prediction of PGD after heart transplant with machine learning (ML). MATERIALS AND METHODS: With the strong power of artificial intelligence, we propose to design a ML algorithm to precisely predict the PGD with the donor and recipient features. Moreover, we apply the computational method to automatically identify important features and interactions between them. RESULTS: To evaluate the effectiveness of the ML algorithm in PGD prediction, we curated a PGD patients' cohort from the United Network for Organ Sharing database, which contains 8008 recipients. 5 commonly used ML models are used for performance comparison. The multi-layer perceptron model achieves superior performance, as measured by area under the receiver operating characteristic curve (AUROC), at 0.868. We identify the top 20 important features and interactions between donors and recipients. Clinical analyses are conducted on the identified features and interactions. DISCUSSION: We summarize the contributions of this work from three aspects including methodology, clinical analysis, and insights. We discuss the limitations of this work on data, model, and real-world implementation perspectives. Additionally, we further discuss the future directions to extend this work to more organ types and diseases. CONCLUSION: In summary, ML has promising applications in PGD prediction as a computational tool for clinical study. We can also use the ML model to help us identify and discover new risk factors and interactions between donor and recipient.

Knowledge graph data are prevalent in real-world applications, and knowledge graph neural networks (KGNNs) are essential techniques for knowledge graph representation learning. Although KGNN effectively models the structural information from knowledge graphs, these frameworks amplify the underlying data bias that leads to discrimination towards certain groups or individuals in resulting applications. Additionally, as existing debiasing approaches mainly focus on entity-wise bias, eliminating the multi-hop relational bias that pervasively exists in knowledge graphs remains an open question. However, it is very challenging to eliminate relational bias due to the sparsity of the paths that generate the bias and the non-linear proximity structure of knowledge graphs. To tackle the challenges, we propose Fair-KGNN, a KGNN framework that simultaneously alleviates multi-hop bias and preserves the proximity information of entity-to-relation in knowledge graphs. The proposed framework is generalizable to mitigate relational bias for all types of KGNN. Fair-KGNN is applicable to incorporate two state-of-the-art KGNN models, RGCN and CompGCN, to mitigate gender-occupation and nationality-salary bias. The experiments carried out on three benchmark knowledge graph datasets demonstrate that Fair-KGNN can effectively mitigate unfair situations during representation learning while preserving the predictive performance of KGNN models. The source code of the proposed method is available at: https://github.com/ynchuang/Mitigating-Relational-Bias-on-Knowledge-Graphs .

As Graph Neural Networks (GNNs) have been widely used in real-world applications, model explanations are required not only by users but also by legal regulations. However, simultaneously achieving high fidelity and low computational costs in generating explanations has been a challenge for current methods. In this work, we propose a framework of GNN explanation named L e A rn R emoval-based A ttribution (LARA) to address this problem. Specifically, we introduce removal-based attribution and demonstrate its substantiated link to interpretability fidelity theoretically and experimentally. The explainer in LARA learns to generate removal-based attribution which enables providing explanations with high fidelity. A strategy of subgraph sampling is designed in LARA to improve the scalability of the training process. In the deployment, LARA can efficiently generate the explanation through a feed-forward pass. We benchmark our approach with other state-of-the-art GNN explanation methods on six datasets. Results highlight the effectiveness of our framework regarding both efficiency and fidelity. In particular, LARA is 3.1 \(\times\) faster and achieves higher fidelity than the state-of-the-art method on the large dataset ogbn-arxiv (more than 160K nodes and 1M edges), showing its great potential in real-world applications. Our source code is available at https://github.com/yaorong0921/LARA .

There has been significant progress in improving the performance of graph neural networks (GNNs) through enhancements in graph data, model architecture design, and training strategies. For fairness in graphs, recent studies achieve fair representations and predictions through either graph data pre-processing (e.g., node feature masking, and topology rewiring) or fair training strategies (e.g., regularization, adversarial debiasing, and fair contrastive learning). How to achieve fairness in graphs from the model architecture perspective is less explored. More importantly, GNNs exhibit worse fairness performance compared to multilayer perception since their model architecture (i.e., neighbor aggregation) amplifies biases. To this end, we aim to achieve fairness via a new GNN architecture. We propose Fair Message Passing (FMP) designed within a unified optimization framework for GNNs. Notably, FMP explicitly renders sensitive attribute usage in forward propagation for node classification task using cross-entropy loss without data pre-processing. In FMP, the aggregation is first adopted to utilize neighbors' information and then the bias mitigation step explicitly pushes demographic group node presentation centers together. In this way, FMP scheme can aggregate useful information from neighbors and mitigate bias to achieve better fairness and prediction tradeoff performance. Experiments on node classification tasks demonstrate that the proposed FMP outperforms several baselines in terms of fairness and accuracy on three real-world datasets. The code is available at https://github.com/zhimengj0326/FMP.

Guanchu Wang, Yu-Neng Chuang, Ruixiang Tang, Shaochen Zhong, Jiayi Yuan, Hongye Jin, Zirui Liu, Vipin Chaudhary, Shuai Xu, James Caverlee, Xia Hu. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

The ongoing challenges in time series anomaly detection (TSAD), including the scarcity of anomaly labels and the variability in anomaly lengths and shapes, have led to the need for a more robust and efficient solution. As limited anomaly labels hinder traditional supervised models in anomaly detection, various state-of-the-art (SOTA) deep learning (DL) techniques (e.g., self-supervised learning) are introduced to tackle this issue. However, they encounter difficulties handling variations in anomaly lengths and shapes, limiting their adaptability to diverse anomalies. Additionally, many benchmark datasets suffer from the problem of having explicit anomalies that even random functions can detect. This problem is exacerbated by an ill-posed evaluation metric, known as point adjustment (PA), which results in inflated model performance. In this context, we propose a novel self-supervised learning based Tri-domain Anomaly Detector (TriAD), which addresses these challenges by modeling features across three aspects - temporal, frequency, and residual domains - without relying on anomaly labels. Unlike traditional contrastive learning methods, TriAD employs both inter-domain and intra-domain contrastive loss to learn common attributes among normal data and differentiate them from anomalies. Additionally, our approach can detect anomalies of varying lengths by integrating with a discord discovery algorithm. It is worth noting that this study is the first to reevaluate the DL potential in TSAD, utilizing both rigorously designed datasets and evaluation metrics. Experimental results demonstrate that TriAD achieves a consistent and significant performance increase over both DL and non-DL SOTA baselines. Moreover, in comparison to SOTA discord discovery algorithms, TriAD improves anomaly detection accuracy by 50 % while cutting the inference time down to just one-tenth. Illuminating the significance of rigorous datasets and evaluation metrics, this paper offers a new direction for addressing the multifaceted challenges of TSAD. The source code is publicly available at https://github.com/pseudo-Skye/TriAD.

This paper introduces the Fair Fairness Benchmark (FFB), a benchmarking framework for in-processing group fairness methods. Ensuring fairness in machine learning is important for ethical compliance. However, there exist challenges in comparing and developing fairness methods due to inconsistencies in experimental settings, lack of accessible algorithmic implementations, and limited extensibility of current fairness packages and tools. To address these issues, we introduce an open-source standardized benchmark for evaluating in-processing group fairness methods and provide a comprehensive analysis of state-of-the-art methods to ensure different notions of group fairness. This work offers the following key contributions: the provision of flexible, extensible, minimalistic, and research-oriented open-source code; the establishment of unified fairness method benchmarking pipelines; and extensive benchmarking, which yields key insights from 45,079 experiments, 14,428 GPU hours. We believe that our work will significantly facilitate the growth and development of the fairness research community. The benchmark is available at https://github.com/ahxt/fair_fairness_benchmark.

It is well known that LLMs cannot generalize well to long contexts whose lengths are larger than the training sequence length. This poses challenges when employing LLMs for processing long input sequences during inference. In this work, we argue that LLMs themselves have inherent capabilities to handles s long contexts without fine-tuning. To achieve this goal, we propose SelfExtend to extend the context window of LLMs by constructing bi-level attention information: the grouped attention and the neighbor attention. The grouped attention captures the dependencies among tokens that are far apart, while neighbor attention captures dependencies among adjacent tokens within a specified range. The two-level attentions are computed based on the original model’s self-attention mechanism during inference. With minor code modification, our SelfExtend can effortlessly extend existing LLMs’ context window without any fine-tuning. We conduct comprehensive experiments on multiple benchmarks and the results show that our SelfExtend can effectively extend existing LLMs’ context window length.

Efficiently serving large language models (LLMs) requires batching many requests together to reduce the cost per request. Yet, the key-value (KV) cache, which stores attention keys and values to avoid re-computations, significantly increases memory demands and becomes the new bottleneck in speed and memory usage. This memory demand increases with larger batch sizes and longer context lengths. Additionally, the inference speed is limited by the size of KV cache, as the GPU's SRAM must load the entire KV cache from the main GPU memory for each token generated, causing the computational core to be idle during this process. A straightforward and effective solution to reduce KV cache size is quantization, which decreases the total bytes taken by KV cache. However, there is a lack of in-depth studies that explore the element distribution of KV cache to understand the hardness and limitation of KV cache quantization. To fill the gap, we conducted a comprehensive study on the element distribution in KV cache of popular LLMs. Our findings indicate that the key cache should be quantized per-channel, i.e., group elements along the channel dimension and quantize them together. In contrast, the value cache should be quantized per-token. From this analysis, we developed a tuning-free 2bit KV cache quantization algorithm, named KIVI. With the hardware-friendly implementation, KIVI can enable Llama (Llama-2), Falcon, and Mistral models to maintain almost the same quality while using 2.6$\times$ less peak memory usage (including the model weight). This reduction in memory usage enables up to 4x larger batch size, bringing $2.35 \times \sim 3.47 \times$ throughput on real LLM inference workload.

Explainable machine learning significantly improves the transparency of deep neural networks. However, existing work is constrained to explaining the behavior of individual model predictions, and lacks the ability to transfer the explanation across various models and tasks. This limitation results in explaining various tasks being time- and resource-consuming. To address this problem, we introduce a Transferable Vision Explainer (TVE) that can effectively explain various vision models in downstream tasks. Specifically, the transferability of TVE is realized through a pre-training process on large-scale datasets towards learning the meta-attribution. This meta-attribution leverages the versatility of generic backbone encoders to comprehensively encode the attribution knowledge for the input instance, which enables TVE to seamlessly transfer to explaining various downstream tasks, without the need for training on task-specific data. Empirical studies involve explaining three different architectures of vision models across three diverse downstream datasets. The experiment results indicate TVE is effective in explaining these tasks without the need for additional training on downstream data.

Model compression is one of the most popular approaches to improve the accessibility of Large Language Models (LLMs) by reducing their memory footprint. However, the gaining of such efficiency benefits often simultaneously demands extensive engineering efforts and intricate designs to mitigate the performance decline. In this work, we leverage *(Soft) Prompt Tuning* in its most vanilla form and discover such conventionally learned soft prompts can recover the performance of compressed LLMs. More surprisingly, we observe such recovery effect to be transferable among different tasks and models (albeit natural tokenizer and dimensionality limitations), resulting in further overhead reduction and yet, subverting the common belief that learned soft prompts are task-specific. Our work is fully orthogonal and compatible with model compression frameworks such as pruning and quantization, where we enable up to $8\times$ compressed LLM (with a joint 4-bit quantization and 50% weight pruning compression) to match its uncompressed counterparts on popular benchmarks. We note that we are the first to reveal vanilla Parameter-Efficient Fine-Tuning (PEFT) techniques have the potential to be utilized under a compression recovery context, opening a new line of opportunities for model accessibility advancement while freeing our fellow researchers from the previously present engineering burdens and constraints. The code is available at https://github.com/zirui-ray-liu/compress-then-prompt.

Suppose a self-driving car is crashing into pedestrians, or a chatbot is instructing its users to conduct criminal wrongdoing; the stakeholders of such products will undoubtedly want to patch these catastrophic errors as soon as possible. To address such concerns, Model Editing: the study of efficiently patching model behaviors without significantly altering their general performance, has seen considerable activity, with hundreds of editing techniques developed in various domains such as CV and NLP. However, the graph learning community has objectively fallen behind with only a few Graph Neural Network-compatible — and just one GNN-specific — model editing methods available, where all of which are limited in their practical scope. We argue that the impracticality of these methods lies in their lack of Sequential Editing Robustness: the ability to edit multiple errors sequentially, and therefore fall short in effectiveness, as this approach mirrors how errors are discovered and addressed in the real world. In this paper, we delve into the specific reasons behind the difficulty of editing GNNs in succession and observe the root cause to be model overfitting. We subsequently propose a simple yet effective solution — SEED-GNN — by leveraging overfit-prevention techniques in a GNN-specific context to derive the first and only GNN model editing method that scales practically. Additionally, we formally frame the task paradigm of GNN editing and hope to inspire future research in this crucial but currently overlooked field. Please refer to our GitHub repository for code and checkpoints.

Yu-Neng Chuang, Tianwei Xing, Chia-Yuan Chang, Zirui Liu, Xun Chen, Xia Hu. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Deep neural networks are ubiquitously adopted in many applications, such as computer vision, natural language processing, and graph analytics. However, well-trained neural networks can make prediction errors after deployment as the world changes. \textit{Model editing} involves updating the base model to correct prediction errors with less accessible training data and computational resources. Despite recent advances in model editors in computer vision and natural language processing, editable training in graph neural networks (GNNs) is rarely explored. The challenge with editable GNN training lies in the inherent information aggregation across neighbors, which can lead model editors to affect the predictions of other nodes unintentionally. In this paper, we first observe the gradient of cross-entropy loss for the target node and training nodes with significant inconsistency, which indicates that directly fine-tuning the base model using the loss on the target node deteriorates the performance on training nodes. Motivated by the gradient inconsistency observation, we propose a simple yet effective \underline{G}radient \underline{R}ewiring method for \underline{E}ditable graph neural network training, named \textbf{GRE}. Specifically, we first store the anchor gradient of the loss on training nodes to preserve the locality. Subsequently, we rewire the gradient of the loss on the target node to preserve performance on the training node using anchor gradient. Experiments demonstrate the effectiveness of GRE on various model architectures and graph datasets in terms of multiple editing situations. The source code is available at \url{https://github.com/zhimengj0326/Gradient_rewiring_editing}.

Contrastive Learning (CL) has shown promising performance in collaborative filtering. The key idea is to use contrastive loss to generate augmentation-invariant embeddings by maximizing the Mutual Information between different augmented views of the same instance. However, we empirically observe that existing CL models suffer from the dimensional collapse issue, where user/item embeddings only span a low-dimension subspace of the entire feature space. This suppresses other dimensional information and weakens the distinguishability of embeddings. Here we propose a non-contrastive learning objective, named nCL, which explicitly mitigates dimensional collapse of representations in collaborative filtering. Our nCL aims to achieve geometric properties of Alignment and Compactness on the embedding space. In particular, the alignment tries to push together representations of positive-related user-item pairs, while compactness tends to find the optimal coding length of user/item embeddings, subject to a given distortion. More importantly, our nCL does not require data augmentation nor negative sampling during training, making it scalable to large datasets compared to contrastive learning methods. Experimental results demonstrate the superiority of our nCL.

This article presents a comprehensive and practical guide for practitioners and end-users working with Large Language Models (LLMs) in their downstream Natural Language Processing (NLP) tasks. We provide discussions and insights into the usage of LLMs from the perspectives of models, data, and downstream tasks. First, we offer an introduction and brief summary of current language models. Then, we discuss the influence of pre-training data, training data, and test data. Most importantly, we provide a detailed discussion about the use and non-use cases of large language models for various natural language processing tasks, such as knowledge-intensive tasks, traditional natural language understanding tasks, generation tasks, emergent abilities, and considerations for specific tasks. We present various use cases and non-use cases to illustrate the practical applications and limitations of LLMs in real-world scenarios. We also try to understand the importance of data and the specific challenges associated with each NLP task. Furthermore, we explore the impact of spurious biases on LLMs and delve into other essential considerations, such as efficiency, cost, and latency, to ensure a comprehensive understanding of deploying LLMs in practice. This comprehensive guide aims to provide researchers and practitioners with valuable insights and best practices for working with LLMs, thereby enabling the successful implementation of these models in a wide range of NLP tasks. A curated list of practical guide resources of LLMs, regularly updated, can be found at https://github.com/Mooler0410/LLMsPracticalGuide . An LLMs evolutionary tree, editable yet regularly updated, can be found at llmtree.ai .

Graph neural networks (GNNs) have shown prominent performance on attributed network embedding. However, existing efforts mainly focus on exploiting network structures, while the exploitation of node attributes is rather limited as they only serve as node features at the initial layer. This simple strategy impedes the potential of node attributes in augmenting node connections, leading to limited receptive field for inactive nodes with few or even no neighbors. Furthermore, the training objectives (i.e., reconstructing network structures) of most GNNs also do not include node attributes, although studies have shown that reconstructing node attributes is beneficial. Thus, it is encouraging to deeply involve node attributes in the key components of GNNs, including graph convolution operations and training objectives. However, this is a nontrivial task since an appropriate way of integration is required to maintain the merits of GNNs. To bridge the gap, in this paper, we propose COllaborative graph Neural Networks–CONN, a tailored GNN architecture for attribute network embedding. It improves model capacity by 1) selectively diffusing messages from neighboring nodes and involved attribute categories, and 2) jointly reconstructing node-to-node and node-to-attribute-category interactions via cross-correlation. Experiments on real-world networks demonstrate that CONN excels state-of-the-art embedding algorithms with a great margin.

The rapid advancements in deep learning technologies have profoundly influenced the field of medical image analysis, yet their full integration into clinical radiology practices has not progressed as quickly as expected. A significant hurdle to their widespread adoption among radiologists and clinicians is the prevailing lack of trust and confidence in the outcomes produced by these technologies. This concern primarily stems from concerns regarding the explainability and generalizability of deep learning models within the realm of medical imaging. As part of the responses from the Medical Image Analysis Community to address these critical issues, we organized the IEEE Transactions on Neural Networks and Learning Systems (TNNLS) Special Issue on explainable and generalizable deep learning for medical imaging. This IEEE TNNLS Special Issue calls for original and innovative methodological contributions that aim to address the key challenges on explainability and generalizability of deep learning for medical imaging. This IEEE TNNLS Special Issue emphasizes the research and advanced development of the technical aspects of new image analysis methodologies, and all the developed new methods should also be evaluated or validated on real and large-scale medical imaging data.

The exponential growth in scholarly publications necessitates advanced tools for efficient article retrieval, especially in interdisciplinary fields where diverse terminologies are used to describe similar research. Traditional keyword-based search engines often fall short in assisting users who may not be familiar with specific terminologies. To address this, we present a knowledge graph based paper search engine for biomedical research to enhance the user experience in discovering relevant queries and articles. The system, dubbed DiscoverPath, employs Named Entity Recognition (NER) and part-of-speech (POS) tagging to extract terminologies and relationships from article abstracts to create a KG. To reduce information overload, DiscoverPath presents users with a focused subgraph containing the queried entity and its neighboring nodes and incorporates a query recommendation system enabling users to iteratively refine their queries. The system is equipped with an accessible Graphical User Interface that provides an intuitive visualization of the KG, query recommendations, and detailed article information, enabling efficient article retrieval, thus fostering interdisciplinary knowledge exploration. DiscoverPath is open-sourced at https://github.com/ynchuang/DiscoverPath with a demo video at Youtube.

Imbalanced datasets are commonly observed in various real-world applications, presenting significant challenges in training classifiers. When working with large datasets, the imbalanced issue can be further exacerbated, making it exceptionally difficult to train classifiers effectively. To address the problem, over-sampling techniques have been developed to linearly interpolating data instances between minorities and their neighbors. However, in many real-world scenarios such as anomaly detection, minority instances are often dispersed diversely in the feature space rather than clustered together. Inspired by domain-agnostic data mix-up, we propose generating synthetic samples iteratively by mixing data samples from both minority and majority classes. It is non-trivial to develop such a framework, the challenges include source sample selection, mix-up strategy selection, and the coordination between the underlying model and mix-up strategies. To tackle these challenges, we formulate the problem of iterative data mix-up as a Markov decision process (MDP) that maps data attributes onto an augmentation strategy. To solve the MDP, we employ an actor-critic framework to adapt the discrete-continuous decision space. This framework is utilized to train a data augmentation policy and design a reward signal that explores classifier uncertainty and encourages performance improvement, irrespective of the classifier's convergence. We demonstrate the effectiveness of our proposed framework through extensive experiments conducted on seven publicly available benchmark datasets using three different types of classifiers. The results of these experiments showcase the potential and promise of our framework in addressing imbalanced datasets with diverse minorities.

The increasing prevalence of Deep Neural Networks (DNNs) in cloud-based services has led to their widespread use through various APIs. However, recent studies reveal the susceptibility of these public APIs to model extraction attacks, where adversaries attempt to create a local duplicate of the private model using data and API-generated predictions. Existing defense methods often involve perturbing prediction distributions to hinder an attacker's training goals, inadvertently affecting API utility. In this study, we extend the concept of digital watermarking to protect DNNs' APIs. We suggest embedding a watermark into the safeguarded APIs; thus, any model attempting to copy will inherently carry the watermark, allowing the defender to verify any suspicious models. We propose a simple yet effective framework to increase watermark transferability. By requiring the model to memorize the preset watermarks in the final decision layers, we significantly enhance the transferability of watermarks. Comprehensive experiments show that our proposed framework not only successfully watermarks APIs but also maintains their utility.

Mengnan Du, Subhabrata Mukherjee, Yu Cheng, Milad Shokouhi, Xia Hu, Ahmed Hassan Awadallah. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023.

The recent breakthrough achieved by graph neural networks (GNNs) with few labeled data accelerates the pace of deploying GNNs on real-world applications. While several efforts have been made to scale GNNs training for large-scale graphs, GNNs still suffer from the scalability challenge of model inference, due to the graph dependency issue incurred by the message-passing mechanism, therefore hindering its deployment in resource-constrained applications. An intuitive remedy is compressing the cumbersome GNN model into inference-friendly multi-layer perceptrons (MLPs) using knowledge distillation (KD). However, the standard KD strategy, i.e., training MLPs using the soft labels of labeled and unlabeled nodes from the teacher, is suboptimal, since the GNN teacher would inevitably make wrong predictions for unlabeled data, especially in the semi-supervised scenario. To address this, we propose a novel Reliable Knowledge Distillation framework for MLP optimization (RKDMLP), which shows strong promise in achieving a “sweet point” in co-optimizing model accuracy and efficiency. Its core insight is to use a meta-policy to filter out those unreliable soft labels. To train the meta-policy, we design a reward-driven objective based on a meta-set and adopt policy gradient to optimize the expected reward. Then we apply the meta-policy to the unlabeled nodes and select the most reliable soft labels for distillation. Extensive experiments across various GNN backbones, on 7 small graphs and 2 large-scale datasets from the challenging Open Graph Benchmark, demonstrate the superiority of our proposal. Moreover, RKD-MLP also shows good robustness iv.r.t. graph topology and node feature noises.

Training graph neural networks (GNNs) is extremely time consuming because sparse graph-based operations are hard to be accelerated by community hardware. Prior art successfully reduces the computation cost of dense matrix based operations (e.g., convolution and linear) via sampling-based approximation. However, unlike dense matrices, sparse matrices are stored in the irregular data format such that each row/column may have different number of non-zero entries. Thus, compared to the dense counterpart, approximating sparse operations has two unique challenges (1) we cannot directly control the efficiency of approximated sparse operation since the computation is only executed on non-zero entries; (2) sampling sparse matrices is much more inefficient due to the irregular data format. To address the issues, our key idea is to control the accuracy-efficiency trade off by optimizing computation resource allocation layer-wisely and epoch-wisely. For the first challenge, we customize the computation resource to different sparse operations, while limit the total used resource below a certain budget. For the second challenge, we cache previous sampled sparse matrices to reduce the epoch-wise sampling overhead. Finally, we propose a switching mechanisms to improve the generalization of GNNs trained with approximated operations. To this end, we propose Randomized Sparse Computation. In practice, rsc can achieve up to 11.6X speedup for a single sparse operation and 1.6X end-to-end wall-clock time speedup with almost no accuracy drop.

Activation compressed training provides a solution towards reducing the memory cost of training deep neural networks (DNNs). However, state-of-the-art work combines a search of quantization bit-width with the training, which makes the procedure complicated and less transparent. To this end, we propose a simple and effective method to compress DNN training. Our method is motivated by an instructive observation: DNN backward propagation mainly utilizes the low-frequency component (LFC) of the activation maps, while the majority of memory is for caching the high-frequency component (HFC) during the training. This indicates the HFC of activation maps is highly redundant and compressible, which inspires our proposed Dual Activation Precision (DIVISION). During the training, DIVISION preserves a high-precision copy of LFC and compresses the HFC into a light-weight copy with low numerical precision. This can significantly reduce the memory cost while maintaining a competitive model accuracy. Experiment results show DIVISION has better comprehensive performance than state-of-the-art methods, including over 10x compression of activation maps and competitive training throughput, without loss of model accuracy. The source code is available at https://github.com/guanchuwang/division.

Transformer-based models are powerful for modeling temporal dynamics of user preference in sequential recommendation. Most of the variants adopt the Softmax transformation in the self-attention layers to generate dense attention probabilities. However, real-world item sequences are often noisy, containing a mixture of true-positive and false-positive interactions. Such dense attentions inevitably assign probability mass to noisy or irrelevant items, leading to sub-optimal performance and poor explainability. Here we propose a Probabilistic Masked Attention Network (PMAN) to identify the sparse pattern of attentions, which is more desirable for pruning noisy items in sequential recommendation. Specifically, we employ a probabilistic mask to achieve sparse attentions under a constrained optimization framework. As such, PMAN allows to select which information is critical to be retained or dropped in a data-driven fashion. Experimental studies on real-world benchmark datasets show that PMAN is able to improve the performance of Transformers significantly.

Feed recommendation systems, which recommend a sequence of items for users to browse and interact with, have gained significant popularity in practical applications. In feed products, users tend to browse a large number of items in succession, so the previously viewed items have a significant impact on users' behavior towards the following items. Therefore, traditional methods that mainly focus on improving the accuracy of recommended items are suboptimal for feed recommendations because they may recommend highly similar items. For feed recommendation, it is crucial to consider both the accuracy and diversity of the recommended item sequences in order to satisfy users' evolving interest when consecutively viewing items. To this end, this work proposes a general re-ranking framework named Multi-factor Sequential Re-ranking with Perception-Aware Diversification~(MPAD) to jointly optimize accuracy and diversity for feed recommendation in a sequential manner. Specifically, MPAD first extracts users' different scales of interests from their behavior sequences through graph clustering-based aggregations. Then, MPAD proposes two sub-models to respectively evaluate the accuracy and diversity of a given item by capturing users' evolving interest due to the ever-changing context and users' personal perception of diversity from an item sequence perspective. This is consistent with the browsing nature of the feed scenario. Finally, MPAD generates the return list by sequentially selecting optimal items from the candidate set to maximize the joint benefits of accuracy and diversity of the entire list. MPAD has been implemented in Taobao's homepage feed to serve the main traffic and provide services to recommend billions of items to hundreds of millions of users every day.

The role of data in AI has been significantly magnified by the emerging concept of data-centric AI. In contrast to the traditional model-centric paradigm, which focuses on developing more effective models given fixed datasets, data-centric AI emphasizes the systematic engineering of data in building AI systems. However, as a new concept, many critical aspects of data-centric AI remain ambiguous, such as its definitions, associated tasks, algorithms, challenges, and benchmarks. This tutorial aims to review and discuss this emerging field, with a particular focus on the three general data-centric AI goals: training data development, inference data development, and data maintenance. The objective of this tutorial is threefold: (1) to formally categorize the field of data-centric AI using a goal-driven taxonomy and discuss the needs and challenges of each goal, (2) to comprehensively review the state-of-the-art techniques, and (3) to discuss the future perspectives and open research directions to inspire further innovations in this field.

As graph neural networks (GNNs) struggle with large-scale graphs due to high computational demands, data distillation for graph data promises to alleviate this issue by distilling a large real graph into a smaller distilled graph while maintaining comparable prediction performance for GNNs trained on both graphs. However, we observe that GNNs trained on distilled graphs may exhibit more severe group fairness problems than those trained on real graphs. Motivated by this observation, we propose \textit{fair graph distillation} (\Algnameabbr), an approach for generating small distilled \textit{fair and informative} graphs based on the graph distillation method. The challenge lies in the deficiency of sensitive attributes for nodes in the distilled graph, making most debiasing methods (e.g., regularization and adversarial debiasing) intractable for distilled graphs. We develop a simple yet effective bias metric, called coherence, for distilled graphs. Based on the proposed coherence metric, we introduce a framework for fair graph distillation using a bi-level optimization algorithm. Extensive experiments demonstrate that the proposed algorithm can achieve better prediction performance-fairness trade-offs across various datasets and GNN architectures.

Fairness in machine learning has attracted increasing attention in recent years. The fairness methods improving algorithmic fairness for in-distribution data may not perform well under distribution shifts. In this paper, we first theoretically demonstrate the inherent connection between distribution shift, data perturbation, and model weight perturbation. Subsequently, we analyze the sufficient conditions to guarantee fairness (i.e., low demographic parity) for the target dataset, including fairness for the source dataset, and low prediction difference between the source and target datasets for each sensitive attribute group. Motivated by these sufficient conditions, we propose robust fairness regularization (RFR) by considering the worst case within the model weight perturbation ball for each sensitive attribute group. We evaluate the effectiveness of our proposed RFR algorithm on synthetic and real distribution shifts across various datasets. Experimental results demonstrate that RFR achieves better fairness-accuracy trade-off performance compared with several baselines. The source code is available at \url{https://github.com/zhimengj0326/RFR_NeurIPS23}.

As the model size grows rapidly, fine-tuning the large pre-trained language model has become increasingly difficult due to its extensive memory usage. Previous works usually focus on reducing the number of trainable parameters in the network. While the model parameters do contribute to memory usage, the primary memory bottleneck during training arises from storing feature maps, also known as activations, as they are crucial for gradient calculation. Notably, machine learning models are typically trained using stochastic gradient descent. We argue that in stochastic optimization, models can handle noisy gradients as long as the gradient estimator is unbiased with reasonable variance. Following this motivation, we propose a new family of unbiased estimators called \sas, for matrix production with reduced variance, which only requires storing the sub-sampled activations for calculating the gradient. Our work provides both theoretical and experimental evidence that, in the context of tuning transformers, our proposed estimators exhibit lower variance compared to existing ones. By replacing the linear operation with our approximated one in transformers, we can achieve up to 2.7X peak memory reduction with almost no accuracy drop and enables up to $6.4\times$ larger batch size. Under the same hardware, \sas enables better down-streaming task performance by applying larger models and/or faster training speed with larger batch sizes. The code is available at https://anonymous.4open.science/r/WTACRS-A5C5/.

In the field of natural language processing, the prevalent approach involves fine-tuning pretrained language models (PLMs) using local samples. Recent research has exposed the susceptibility of PLMs to backdoor attacks, wherein the adversaries can embed malicious prediction behaviors by manipulating a few training samples. In this study, our objective is to develop a backdoor-resistant tuning procedure that yields a backdoor-free model, no matter whether the fine-tuning dataset contains poisoned samples. To this end, we propose and integrate an \emph{honeypot module} into the original PLM, specifically designed to absorb backdoor information exclusively. Our design is motivated by the observation that lower-layer representations in PLMs carry sufficient backdoor features while carrying minimal information about the original tasks. Consequently, we can impose penalties on the information acquired by the honeypot module to inhibit backdoor creation during the fine-tuning process of the stem network. Comprehensive experiments conducted on benchmark datasets substantiate the effectiveness and robustness of our defensive strategy. Notably, these results indicate a substantial reduction in the attack success rate ranging from 10\% to 40\% when compared to prior state-of-the-art methods.

Densely structured pruning methods utilizing simple pruning heuristics can deliver immediate compression and acceleration benefits with acceptable benign performances. However, empirical findings indicate such naively pruned networks are extremely fragile under simple adversarial attacks. Naturally, we would be interested in knowing if such a phenomenon also holds for carefully designed modern structured pruning methods. If so, then to what extent is the severity? And what kind of remedies are available? Unfortunately, both the questions and the solution remain largely unaddressed: no prior art is able to provide a thorough investigation on the adversarial performance of modern structured pruning methods (spoiler: it is not good), yet the few works that attempt to provide mitigation often do so at various extra costs with only to-be-desired performance. In this work, we answer both questions by fairly and comprehensively investigating the adversarial performance of 10+ popular structured pruning methods. Solution-wise, we take advantage of *Grouped Kernel Pruning (GKP)*'s recent success in pushing densely structured pruning freedom to a more fine-grained level. By mixing up kernel smoothness — a classic robustness-related kernel-level metric — into a modified GKP procedure, we present a one-shot-post-train-weight-dependent GKP method capable of advancing SOTA performance on both the benign and adversarial scale, while requiring no extra (in fact, often less) cost than a standard pruning procedure. Please refer to our [GitHub repository](https://github.com/henryzhongsc/adv_robust_gkp) for code implementation, tool sharing, and model checkpoints.

Graph Neural Networks (GNNs) have achieved state-of-the-art performance in recommender systems. Nevertheless, the process of searching and ranking from a large item corpus usually requires high latency, which limits the widespread deployment of GNNs in industry-scale applications. To address this issue, many methods compress user/item representations into the binary embedding space to reduce space requirements and accelerate inference. Also, they use the Straight-through Estimator (STE) to prevent vanishing gradients during back-propagation. However, the STE often causes the gradient mismatch problem, leading to sub-optimal results.

Time series anomaly detection is a challenging task with a wide range of real-world applications. Due to label sparsity, training a deep anomaly detector often relies on unsupervised approaches. Recent efforts have been devoted to time series domain adaptation to leverage knowledge from similar domains. However, existing solutions may suffer from negative knowledge transfer on anomalies due to their diversity and sparsity. Motivated by the empirical study of context alignment between two domains, we aim to transfer knowledge between two domains via adaptively sampling context information for two domains. This is challenging because it requires simultaneously modeling the complex in-domain temporal dependencies and cross-domain correlations while exploiting label information from the source domain. To this end, we propose a framework that combines context sampling and anomaly detection into a joint learning procedure. We formulate context sampling into the Markov decision process and exploit deep reinforcement learning to optimize the time series domain adaptation process via context sampling and design a tailored reward function to generate domain-invariant features that better align two domains for anomaly detection. Experiments on three public datasets show promise for knowledge transfer between two similar domains and two entirely different domains.

The role of data in building AI systems has recently been significantly magnified by the emerging concept of data-centric AI (DCAI), which advocates a fundamental shift from model advancements to ensuring data quality and reliability. Although our community has continuously invested efforts into enhancing data in different aspects, they are often isolated initiatives on specific tasks. To facilitate the collective initiative in our community and push forward DCAI, we draw a big picture and bring together three general missions: training data development, evaluation data development, and data maintenance. We provide a top-level discussion on representative DCAI tasks and share perspectives. Finally, we list open challenges to motivate future exploration.

Graph neural networks (GNNs), which learn the node representations by recursively aggregating information from its neighbors, have become a predominant computational tool in many domains. To handle large-scale graphs, most of the existing methods partition the input graph into multiple sub-graphs (e.g., through node clustering) and apply batch training to save memory cost. However, such batch training will lead to label bias within each batch and result in over- confidence in model predictions. Since the connected nodes with positively related labels tend to be assigned together, the traditional cross-entropy minimization process attends on the predictions of biased classes at a batch to intensify the overfitting issue. To overcome the problem of label bias, we propose adaptive label smoothing (ALS) method to replace the one-hot hard labels with smoothed ones, which learns to allocate label confidences from the biased classes to the others. Specifically, ALS propagates node labels to aggregate the neighborhood label distribution in a pre-processing step, and then updates the optimal smoothed labels online to adapt to specific graph structure. Experiments on the real-world datasets demonstrate that ALS can be generally applied to the main scalable learning frameworks to calibrate the biased labels and improve generalization performances.

Self-supervised learning (SSL) has been demonstrated to be effective in pre-training models that can be generalized to various downstream tasks. Graph Autoencoder (GAE), an increasingly popular SSL approach on graphs, has been widely explored to learn node representations without ground-truth labels. However, recent studies show that existing GAE methods could only perform well on link prediction tasks, while their performance on classification tasks is rather limited. This limitation casts doubt on the generalizability and adoption of GAE. In this paper, for the first time, we show that GAE can generalize well to both link prediction and classification scenarios, including node-level and graph-level tasks, by redesigning its critical building blocks from the graph masking perspective. Our proposal is called Self-Supervised Graph Autoencoder--S2GAE, which unleashes the power of GAEs with minimal yet nontrivial efforts. Specifically, instead of reconstructing the whole input structure, we randomly mask a portion of edges and learn to reconstruct these missing edges with an effective masking strategy and an expressive decoder network. Moreover, we theoretically prove that S2GAE could be regarded as an edge-level contrastive learning framework, providing insights into why it generalizes well. Empirically, we conduct extensive experiments on 21 benchmark datasets across link prediction and node & graph classification tasks. The results validate the superiority of S2GAE against state-of-the-art generative and contrastive methods. This study demonstrates the potential of GAE as a universal representation learner on graphs. Our code is publicly available at https://github.com/qiaoyu-tan/S2GAE.

Graph neural networks (GNNs) have received remarkable success in link prediction (GNNLP) tasks. Existing efforts first predefine the subgraph for the whole dataset and then apply GNNs to encode edge representations by leveraging the neighborhood structure induced by the fixed subgraph. The prominence of GNNLP methods significantly relies on the adhoc subgraph. Since node connectivity in real-world graphs is complex, one shared subgraph is limited for all edges. Thus, the choices of subgraphs should be personalized to different edges. However, performing personalized subgraph selection is nontrivial since the potential selection space grows exponentially to the scale of edges. Besides, the inference edges are not available during training in link prediction scenarios, so the selection process needs to be inductive. To bridge the gap, we introduce a Personalized Subgraph Selector (PS2) as a plug-and-play framework to automatically, personally, and inductively identify optimal subgraphs for different edges when performing GNNLP. PS2 is instantiated as a bi-level optimization problem that can be efficiently solved differently. Coupling GNNLP models with PS2, we suggest a brand-new angle towards GNNLP training: by first identifying the optimal subgraphs for edges; and then focusing on training the inference model by using the sampled subgraphs. Comprehensive experiments endorse the effectiveness of our proposed method across various GNNLP backbones (GCN, GraphSage, NGCF, LightGCN, and SEAL) and diverse benchmarks (Planetoid, OGB, and Recommendation datasets). Our code is publicly available at https://github.com/qiaoyu-tan/PS2

Training deep graph neural networks (GNNs) is notoriously hard. Besides the standard plights in training deep architectures such as vanishing gradients and overfitting, it also uniquely suffers from over-smoothing, information squashing, and so on, which limits their potential power for encoding the high-order neighbor structure in large-scale graphs. Although numerous efforts are proposed to address these limitations, such as various forms of skip connections, graph normalization, and random dropping, it is difficult to disentangle the advantages brought by a deep GNN architecture from those "tricks" necessary to train such an architecture. Moreover, the lack of a standardized benchmark with fair and consistent experimental settings poses an almost insurmountable obstacle to gauge the effectiveness of new mechanisms. In view of those, we present the first fair and reproducible benchmark dedicated to assessing the "tricks" of training deep GNNs. We categorize existing approaches, investigate their hyperparameter sensitivity, and unify the basic configuration. Comprehensive evaluations are then conducted on tens of representative graph datasets including the recent large-scale Open Graph Benchmark, with diverse deep GNN backbones. We demonstrate that an organic combo of initial connection, identity mapping, group and batch normalization attains the new state-of-the-art results for deep GNNs on large datasets. Codes are available: https://github.com/VITA-Group/Deep_GCN_Benchmarking.

Graph neural networks (GNNs) have received tremendous attention due to their superiority in learning node representations. These models rely on message passing and feature transformation functions to encode the structural and feature information from neighbors. However, stacking more convolutional layers significantly decreases the performance of GNNs. Most recent studies attribute this limitation to the over-smoothing issue, where node embeddings converge to indistinguishable vectors. Through a number of experimental observations, we argue that the main factor degrading the performance is the unstable forward normalization and backward gradient resulted from the improper design of the feature transformation, especially for shallow GNNs where the over-smoothing has not happened. Therefore, we propose a novel orthogonal feature transformation, named Ortho-GConv, which could generally augment the existing GNN backbones to stabilize the model training and improve the model's generalization performance. Specifically, we maintain the orthogonality of the feature transformation comprehensively from three perspectives, namely hybrid weight initialization, orthogonal transformation, and orthogonal regularization. By equipping the existing GNNs (e.g. GCN, JKNet, GCNII) with Ortho-GConv, we demonstrate the generality of the orthogonal feature transformation to enable stable training, and show its effectiveness for node and graph classification tasks.

Recent studies indicate that deep neural networks (DNNs) are prone to show discrimination towards certain demographic groups. We observe that algorithmic discrimination can be explained by the high reliance of the models on fairness sensitive features. Motivated by this observation, we propose to achieve fairness by suppressing the DNN models from capturing the spurious correlation between those fairness sensitive features with the underlying task. Specifically, we firstly train a bias-only teacher model which is explicitly encouraged to maximally employ fairness sensitive features for prediction. The teacher model then counter-teaches a debiased student model so that the interpretation of the student model is orthogonal to the interpretation of the teacher model. The key idea is that since the teacher model relies explicitly on fairness sensitive features for prediction, the orthogonal interpretation loss enforces the student network to reduce its reliance on sensitive features and instead capture more task relevant features for prediction. Experimental analysis indicates that our framework substantially reduces the model's attention on fairness sensitive features. Experimental results on four datasets further validate that our framework has consistently improved the fairness with respect to three group fairness metrics, with a comparable or even better accuracy.

Deep learning models have achieved exceptional predictive performance in a wide variety of tasks, ranging from computer vision, natural language processing, to graph mining. Many businesses and organizations across diverse domains are now building large-scale applications based on deep learning. However, there are growing concerns, regarding the fairness, security, and trustworthiness of these models, largely due to the opaque nature of their decision processes. Recently, there has been an increasing interest in explainable deep learning that aims to reduce the opacity of a model by explaining its behavior, its predictions, or both, thus building trust between human and complex deep learning models. A collection of explanation methods have been proposed in recent years that address the problem of low explainability and opaqueness of models. In this tutorial, we introduce recent explanation methods from a data perspective, targeting models that process image data, text data, and graph data, respectively. We will compare their strengths and limitations, and offer real-world applications.

Deploying deep neural networks (DNNs) on edge devices provides efficient and effective solutions for the real-world tasks. Edge devices have been used for collecting a large volume of data efficiently in different domains. DNNs have been an effective tool for data processing and analysis. However, designing DNNs on edge devices is challenging due to the limited computational resources and memory. To tackle this challenge, we demonstrate oBject detection system for Edge Devices (BED) on the MAX78000 DNN accelerator. It integrates on-device DNN inference with a camera and an LCD display for image acquisition and detection exhibition, respectively. BED is a concise, effective and detailed solution, including model training, quantization, synthesis and deployment. The entire repository is open-sourced on Github1, including a Graphical User Interface (GUI) for on-chip debugging. Experiment results indicate that BED can produce accurate detection with a 300-KB tiny DNN model, which takes only 91.9 ms of inference time and 1.845 mJ of energy. The real-time detection is available at YouTube.

Imbalanced learning is a fundamental challenge in data mining, where there is a disproportionate ratio of training samples in each class. Over-sampling is an effective technique to tackle imbalanced learning through generating synthetic samples for the minority class. While numerous over-sampling algorithms have been proposed, they heavily rely on heuristics, which could be sub-optimal since we may need different sampling strategies for different datasets and base classifiers, and they cannot directly optimize the performance metric. Motivated by this, we investigate developing a learning-based over-sampling algorithm to optimize the classification performance, which is a challenging task because of the huge and hierarchical decision space. At the high level, we need to decide how many synthetic samples to generate. At the low level, we need to determine where the synthetic samples should be located, which depends on the high-level decision since the optimal locations of the samples may differ for different numbers of samples. To address the challenges, we propose AutoSMOTE, an automated over-sampling algorithm that can jointly optimize different levels of decisions. Motivated by the success of SMOTE and its extensions, we formulate the generation process as a Markov decision process (MDP) consisting of three levels of policies to generate synthetic samples within the SMOTE search space. Then we leverage deep hierarchical reinforcement learning to optimize the performance metric on the validation data. Extensive experiments on six real-world datasets demonstrate that AutoSMOTE significantly outperforms the state-of-the-art resampling algorithms. The code is at https://github.com/daochenzha/autosmote

This work develops mixup for graph data. Mixup has shown superiority in improving the generalization and robustness of neural networks by interpolating features and labels between two random samples. Traditionally, Mixup can work on regular, grid-like, and Euclidean data such as image or tabular data. However, it is challenging to directly adopt Mixup to augment graph data because different graphs typically: 1) have different numbers of nodes; 2) are not readily aligned; and 3) have unique typologies in non-Euclidean space. To this end, we propose G-Mixup to augment graphs for graph classification by interpolating the generator (i.e., graphon) of different classes of graphs. Specifically, we first use graphs within the same class to estimate a graphon. Then, instead of directly manipulating graphs, we interpolate graphons of different classes in the Euclidean space to get mixed graphons, where the synthetic graphs are generated through sampling based on the mixed graphons. Extensive experiments show that G-Mixup substantially improves the generalization and robustness of GNNs.

Even though Shapley value provides an effective explanation for a DNN model prediction, the computation relies on the enumeration of all possible input feature coalitions, which leads to the exponentially growing complexity. To address this problem, we propose a novel method SHEAR to significantly accelerate the Shapley explanation for DNN models, where only a few coalitions of input features are involved in the computation. The selection of the feature coalitions follows our proposed Shapley chain rule to minimize the absolute error from the ground-truth Shapley values, such that the computation can be both efficient and accurate. To demonstrate the effectiveness, we comprehensively evaluate SHEAR across multiple metrics including the absolute error from the ground-truth Shapley value, the faithfulness of the explanations, and running speed. The experimental results indicate SHEAR consistently outperforms state-of-the-art baseline methods across different evaluation metrics, which demonstrates its potentials in real-world applications where the computational resource is limited.

Action recognition is an important task for video understanding with broad applications. However, developing an effective action recognition solution often requires extensive engineering efforts in building and testing different combinations of the modules and their hyperparameters. In this demo, we present AutoVideo, a Python system for automated video action recognition. AutoVideo is featured for 1) highly modular and extendable infrastructure following the standard pipeline language, 2) an exhaustive list of primitives for pipeline construction, 3) data-driven tuners to save the efforts of pipeline tuning, and 4) easy-to-use Graphical User Interface (GUI). AutoVideo is released under MIT license at https://github.com/datamllab/autovideo

Learning useful interactions between input features is crucial for tabular data modeling. Recent efforts start to explicitly model the feature interactions with graph, where each feature is treated as an individual node. However, the existing graph construction methods either heuristically formulate a fixed feature-interaction graph based on specific domain knowledge, or simply apply attention function to compute the pairwise feature similarities for each sample. While the fixed graph may be sub-optimal to downstream tasks, the sample-wise graph construction is time-consuming during model training and inference. To tackle these issues, we propose a framework named Table2Graph to transform the feature interaction modeling to learning a unified graph. Represented as a probability adjacency matrix, the unified graph learns to model the key feature interactions shared by the diverse samples in the tabular data. To well optimize the unified graph, we employ the reinforcement learning policy to capture the key feature interactions stably. A sparsity constraint is also proposed to regularize the learned graph from being overly-sparse/smooth. The experimental results in a variety of real-world applications demonstrate the effectiveness and efficiency of our Table2Graph, in terms of the prediction accuracy and feature interaction detection.

Promising progress has been made toward learning efficient time series representations in recent years, but the learned representations often lack interpretability and do not encode semantic meanings by the complex interactions of many latent factors. Learning representations that disentangle these latent factors can bring semantic-rich representations of time series and further enhance interpretability. However, directly adopting the sequential models, such as Long Short-Term Memory Variational AutoEncoder (LSTM-VAE), would encounter a Kullback?Leibler (KL) vanishing problem: the LSTM decoder often generates sequential data without efficiently using latent representations, and the latent spaces sometimes could even be independent of the observation space. And traditional disentanglement methods may intensify the trend of KL vanishing along with the disentanglement process, because they tend to penalize the mutual information between the latent space and the observations. In this paper, we propose Disentangle Time-Series, a novel disentanglement enhancement framework for time series data. Our framework achieves multi-level disentanglement by covering both individual latent factors and group semantic segments. We propose augmenting the original VAE objective by decomposing the evidence lower-bound and extracting evidence linking factorial representations to disentanglement. Additionally, we introduce a mutual information maximization term between the observation space to the latent space to alleviate the KL vanishing problem while preserving the disentanglement property. Experimental results on five real-world IoT datasets demonstrate that the representations learned by DTS achieve superior performance in various tasks with better interpretability.

Embedding learning is an important technique in deep recommendation models to map categorical features to dense vectors. However, the embedding tables often demand an extremely large number of parameters, which become the storage and efficiency bottlenecks. Distributed training solutions have been adopted to partition the embedding tables into multiple devices. However, the embedding tables can easily lead to imbalances if not carefully partitioned. This is a significant design challenge of distributed systems named embedding table sharding, i.e., how we should partition the embedding tables to balance the costs across devices, which is a non-trivial task because 1) it is hard to efficiently and precisely measure the cost, and 2) the partition problem is known to be NP-hard. In this work, we introduce our novel practice in Meta, namely AutoShard, which uses a neural cost model to directly predict the multi-table costs and leverages deep reinforcement learning to solve the partition problem. Experimental results on an open-sourced large-scale synthetic dataset and Meta's production dataset demonstrate the superiority of AutoShard over the heuristics. Moreover, the learned policy of AutoShard can transfer to sharding tasks with various numbers of tables and different ratios of the unseen tables without any fine-tuning. Furthermore, AutoShard can efficiently shard hundreds of tables in seconds. The effectiveness, transferability, and efficiency of AutoShard make it desirable for production use. Our algorithms have been deployed in Meta production environment. A prototype is available at https://github.com/daochenzha/autoshard

Large-scale graph training is a notoriously challenging problem for graph neural networks (GNNs). Due to the nature of evolving graph structures into the training process, vanilla GNNs usually fail to scale up, limited by the GPU memory space. Up to now, though numerous scalable GNN architectures have been proposed, we still lack a comprehensive survey and fair benchmark of this reservoir to find the rationale for designing scalable GNNs. To this end, we first systematically formulate the representative methods of large-scale graph training into several branches and further establish a fair and consistent benchmark for them by a greedy hyperparameter searching. In addition, regarding efficiency, we theoretically evaluate the time and space complexity of various branches and empirically compare them w.r.t GPU memory usage, throughput, and convergence. Furthermore, We analyze the pros and cons for various branches of scalable GNNs and then present a new ensembling training manner, named EnGCN, to address the existing issues. Remarkably, our proposed method has achieved new state-of-the-art (SOTA) performance on large-scale datasets. Our code is available at https://github.com/VITA-Group/LargeScaleGCN_Benchmarking.

We study embedding table placement for distributed recommender systems, which aims to partition and place the tables on multiple hardware devices (e.g., GPUs) to balance the computation and communication costs. Although prior work has explored learning-based approaches for the device placement of computational graphs, embedding table placement remains to be a challenging problem because of 1) the operation fusion of embedding tables, and 2) the generalizability requirement on unseen placement tasks with different numbers of tables and/or devices. To this end, we present DreamShard, a reinforcement learning (RL) approach for embedding table placement. DreamShard achieves the reasoning of operation fusion and generalizability with 1) a cost network to directly predict the costs of the fused operation, and 2) a policy network that is efficiently trained on an estimated Markov decision process (MDP) without real GPU execution, where the states and the rewards are estimated with the cost network. Equipped with sum and max representation reductions, the two networks can directly generalize to any unseen tasks with different numbers of tables and/or devices without fine-tuning. Extensive experiments show that DreamShard substantially outperforms the existing human expert and RNN-based strategies with up to 19% speedup over the strongest baseline on large-scale synthetic tables and our production tables. The code is available.

There has been an explosion of interest in designing various Knowledge Graph Neural Networks (KGNNs), which achieve state-of-the-art performance and provide great explainability for recommendation. The promising performance is mainly resulting from their capability of capturing high-order proximity messages over the knowledge graphs. However, training KGNNs at scale is challenging due to the high memory usage. In the forward pass, the automatic differentiation engines (e.g., TensorFlow/PyTorch) generally need to cache all intermediate activation maps in order to compute gradients in the backward pass, which leads to a large GPU memory footprint. Existing work solves this problem by utilizing multi-GPU distributed frameworks. Nonetheless, this poses a practical challenge when seeking to deploy KGNNs in memory-constrained environments, especially for industry-scale graphs.

We study time-series classification (TSC), a fundamental task of time-series data mining. Prior work has approached TSC from two major directions: (1) similarity-based methods that classify time-series based on the nearest neighbors, and (2) deep learning models that directly learn the representations for classification in a data-driven manner. Motivated by the different working mechanisms within these two research lines, we aim to connect them in such a way as to jointly model time-series similarities and learn the representations. This is a challenging task because it is unclear how we should efficiently leverage similarity information. To tackle the challenge, we propose Similarity-Aware Time-Series Classification (SimTSC), a conceptually simple and general framework that models similarity information with graph neural networks (GNNs). Specifically, we formulate TSC as a node classification problem in graphs, where the nodes correspond to time-series, and the links correspond to pair-wise similarities. We further design a graph construction strategy and a batch training algorithm with negative sampling to improve training efficiency. We instantiate SimTSC with ResNet as the backbone and Dynamic Time Warping (DTW) as the similarity measure. Extensive experiments on the full UCR datasets and several multivariate datasets demonstrate the effectiveness of incorporating similarity information into deep learning models in both supervised and semi-supervised settings. Our code is available at https://github.com/daochenzha/SimTSC.

Graph Neural Networks (GNNs) provide a class of powerful architectures that are effective for graph-based collaborative filtering. Nevertheless, GNNs are known to be vulnerable to adversarial perturbations. Adversarial training is a simple yet effective way to improve the robustness of neural models. For example, many prior studies inject adversarial perturbations into either node features or hidden layers of GNNs. However, perturbing graph structures has been far less studied in recommendations.

Learning discriminative node representations benefits various downstream tasks in graph analysis such as community detection and node classification. Existing graph representation learning methods (e.g., based on random walk and contrastive learning) are limited to maximizing the local similarity of connected nodes. Such pair-wise learning schemes could fail to capture the global distribution of representations, since it has no explicit constraints on the global geometric properties of representation space. To this end, we propose Geometric Graph Representation Learning (G2R) to learn node representations in an unsupervised manner via maximizing rate reduction. In this way, G2R maps nodes in distinct groups (implicitly stored in the adjacency matrix) into different subspaces, while each subspace is compact and different subspaces are dispersedly distributed. G2R adopts a graph neural network as the encoder and maximizes the rate reduction with the adjacency matrix. Furthermore, we theoretically and empirically demonstrate that rate reduction maximization is equivalent to maximizing the principal angles between different subspaces. Experiments on real-world datasets show that G2R outperforms various baselines on node classification and community detection tasks.

Dimensionality reduction is a crucial first step for many unsupervised learning tasks including anomaly detection and clustering. Autoencoder is a popular mechanism to accomplish dimensionality reduction. In order to make dimensionality reduction effective for high-dimensional data embedding nonlinear low-dimensional manifold, it is understood that some sort of geodesic distance metric should be used to discriminate the data samples. Inspired by the success of geodesic distance approximators such as ISOMAP, we propose to use a minimum spanning tree (MST), a graph-based algorithm, to approximate the local neighborhood structure and generate structure-preserving distances among data points. We use this MST-based distance metric to replace the euclidean distance metric in the embedding function of autoencoders and develop a new graph regularized autoencoder, which outperforms a wide range of alternative methods over 20 benchmark anomaly detection datasets. We further incorporate the MST regularizer into two generative adversarial networks and find that using the MST regularizer improves the performance of anomaly detection substantially for both generative adversarial networks. We also test our MST regularized autoencoder on two datasets in a clustering application and witness its superior performance as well.

Over the last decade, deep neural networks (DNNs) are regarded as black-box methods, and their decisions are criticized for the lack of explainability. Existing attempts based on local explanations offer each input a visual saliency map, where the supporting features that contribute to the decision are emphasized with high relevance scores. In this paper, we improve the saliency map based on differentiated explanations, of which the saliency map not only distinguishes the supporting features from backgrounds but also shows the different degrees of importance of the various parts within the supporting features. To do this, we propose to learn a differentiated relevance estimator called DRE, where a carefully-designed distribution controller is introduced to guide the relevance scores towards right-skewed distributions. DRE can be directly optimized under pure classification losses, enabling higher faithfulness of explanations and avoiding non-trivial hyper-parameter tuning. The experimental results on three real-world datasets demonstrate that our differentiated explanations significantly improve the faithfulness with high explainability. Our code and trained models are available at https://github.com/fuweijie/DRE.

Deep models are commonly treated as black-boxes and lack interpretability. Here, we propose a novel approach to interpret deep image classifiers by generating discrete masks. Our method follows the generative adversarial network formalism. The deep model to be interpreted is the discriminator while we train a generator to explain it. The generator is trained to capture discriminative image regions that should convey the same or similar meaning as the original image from the model's perspective. It produces a probability map from which a discrete mask can be sampled. Then the discriminator is used to measure the quality of the sampled mask and provide feedbacks for updating. Due to the sampling operations, the generator cannot be trained directly by back-propagation. We propose to update it using policy gradient. Furthermore, we propose to incorporate gradients as auxiliary information to reduce the search space and facilitate training. We conduct both quantitative and qualitative experiments on the ILSVRC dataset. Experimental results indicate that our method can provide reasonable explanations for predictions and outperform existing approaches. In addition, our method can pass the model randomization test, indicating that it is reasoning the attribution of network predictions.

Embeddings of textual data containing location names (e.g., social media posts) have essential applications in various contexts such as marketing and disaster management. In these downstream implementations, social biases behind location names are highly prone to introduce unfair results through their embeddings; for example, emergent text messages with swapped location names might result in varied rescue responses. Hence, it is critical to address social biases encoded in location names and to seek its mitigation. Prevalent works addressing biases in embeddings mainly focus on individual attributes like gender or ethnicity. Yet, a large number of social attributes behind location names (e.g., income level and population density) makes it challenging to originate the source of biases. Existing mitigation methods based on finding attribute subspaces cannot be simply applied to address social biases. Moreover, bias mitigation tends to simultaneously remove necessary semantics from embeddings, making it difficult to achieve a balance between mitigation performance and semantics retention. In this article, we first employ the concept of counterfactual fairness to investigate the social biases encoded in training data. Then, we quantify the biases in the contextual embeddings (BERT and ELMo). We report a high correlation between biases in the training data and embeddings. Next, we introduce a novel bias mitigation algorithm that customizes bias representations for any location names. The method yields debiased location name vectors for various social attributes simultaneously. The proposed algorithm achieves a better mitigation performance on overall attributes compared with a prevalent postprocessing method, while maintaining correctness by retaining semantic information.

With the prevalence of online social platforms, social recommendation has emerged as a promising direction that leverages the social network among users to enhance recommendation performance. However, the available social relations among users are usually extremely sparse and noisy, which may lead to inferior recommendation performance. To alleviate this problem, this paper novelly exploits the implicit higher-order social influence and dependencies among users to enhance social recommendation. In this paper, we propose a novel embedding method for general bipartite graphs, which defines inter-class message passing between explicit relations and intra-class message passing between implicit higher-order relations via a novel sequential modelling paradigm. Inspired by recent advances in self-attention-based sequential modelling, the proposed model features a self-attentive representation learning mechanism for implicit user-user relations. Moreover, this paper also explores the inductive embedding learning for social recommendation problems to improve the recommendation performance in cold-start settings. The proposed inductive learning paradigm for social recommendation enables embedding inference for those cold-start users and items (unseen during training) as long as they are linked to existing nodes in the original network. Extensive experiments on real-world datasets demonstrate the superiority of our method and suggest that higher-order implicit relationship among users is beneficial to improving social recommendation.

Neural architecture search (NAS) is gaining more and more attention in recent years because of its flexibility and remarkable capability to reduce the burden of neural network design. To achieve better performance, however, the searching process usually costs massive computations that might not be affordable for researchers and practitioners. Although recent attempts have employed ensemble learning methods to mitigate the enormous computational cost, however, they neglect a key property of ensemble methods, namely diversity, which leads to collecting more similar subarchitectures with potential redundancy in the final design. To tackle this problem, we propose a pruning method for NAS ensembles called " subarchitecture ensemble pruning in neural architecture search (SAEP)." It targets to leverage diversity and to achieve subensemble architectures at a smaller size with comparable performance to ensemble architectures that are not pruned. Three possible solutions are proposed to decide which subarchitectures to prune during the searching process. Experimental results exhibit the effectiveness of the proposed method by largely reducing the number of subarchitectures without degrading the performance.

We propose a general and scalable global optimization framework directly operating on annotated graph data by introducing a Bayesian graph neural network to approximate the expensive-to-evaluate objectives. It prevents the cubical complexity of Gaussian processes and can scale linearly with the number of observations. Its parallelized variant makes it scalable. We provide strict theoretical support on its convergence. Intensive experiments conducted on both artificial and real-world problems, including molecular discovery and urban road network design, demonstrate the effectiveness of the proposed methods compared with the current state of the art.

Anomaly detection is an important data mining task with numerous applications, such as intrusion detection, credit card fraud detection, and video surveillance. However, given a specific complicated task with complicated data, the process of building an effective deep learning-based system for anomaly detection still highly relies on human expertise and laboring trials. Also, while neural architecture search (NAS) has shown its promise in discovering effective deep architectures in various domains, such as image classification, object detection, and semantic segmentation, contemporary NAS methods are not suitable for anomaly detection due to the lack of intrinsic search space, unstable search process, and low sample efficiency. To bridge the gap, in this article, we propose AutoAD, an automated anomaly detection framework, which aims to search for an optimal neural network model within a predefined search space. Specifically, we first design a curiosity-guided search strategy to overcome the curse of local optimality. A controller, which acts as a search agent, is encouraged to take actions to maximize the information gain about the controller's internal belief. We further introduce an experience replay mechanism based on self-imitation learning to improve the sample efficiency. Experimental results on various real-world benchmark datasets demonstrate that the deep model identified by AutoAD achieves the best performance, comparing with existing handcrafted models and traditional search methods.

Attribution methods have been developed to understand the decision making process of machine learning models, especially deep neural networks, by assigning importance scores to individual features. Existing attribution methods often built upon empirical intuitions and heuristics. There still lacks a general and theoretical framework that not only can unify these attribution methods, but also theoretically reveal their rationales, fidelity, and limitations. To bridge the gap, in this paper, we propose a Taylor attribution framework and reformulate seven mainstream attribution methods into the framework. Based on reformulations, we analyze the attribution methods in terms of rationale, fidelity, and limitation. Moreover, We establish three principles for a good attribution in the Taylor attribution framework, i.e., low approximation error, correct contribution assignment, and unbiased baseline selection. Finally, we empirically validate the Taylor reformulations, and reveal a positive correlation between the attribution performance and the number of principles followed by the attribution method via benchmarking on real-world datasets.

We present TODS, an automated Time Series Outlier Detection System for research and industrial applications. TODS is a highly modular system that supports easy pipeline construction. The basic building block of TODS is primitive, which is an implementation of a function with hyperparameters. TODS currently supports 70 primitives, including data processing, time series processing, feature analysis, detection algorithms, and a reinforcement module. Users can freely construct a pipeline using these primitives and perform end- to-end outlier detection with the constructed pipeline. TODS provides a Graphical User Interface (GUI), where users can flexibly design a pipeline with drag-and-drop. Moreover, a data-driven searcher is provided to automatically discover the most suitable pipelines given a dataset. TODS is released under Apache 2.0 license at https://github.com/datamllab/tods. A video is available on YouTube (https://youtu.be/JOtYxTclZgQ)

Sequential recommendation has become increasingly essential in various online services. It aims to model the dynamic preferences of users from their historical interactions and predict their next items. The accumulated user behavior records on real systems could be very long. This rich data brings opportunities to track actual interests of users. Prior efforts mainly focus on making recommendations based on relatively recent behaviors. However, the overall sequential data may not be effectively utilized, as early interactions might affect users' current choices. Also, it has become intolerable to scan the entire behavior sequence when performing inference for each user, since real-world system requires short response time. To bridge the gap, we propose a novel long sequential recommendation model, called Dynamic Memory-based Attention Network (DMAN). It segments the overall long behavior sequence into a series of sub-sequences, then trains the model and maintains a set of memory blocks to preserve long-term interests of users. To improve memory fidelity, DMAN dynamically abstracts each user's long-term interest into its own memory blocks by minimizing an auxiliary reconstruction loss. Based on the dynamic memory, the user's short-term and long-term interests can be explicitly extracted and combined for efficient joint recommendation. Empirical results over four benchmark datasets demonstrate the superiority of our model in capturing long-term dependency over various state-of-the-art sequential models.

Human-designed data augmentation strategies have been replaced by automatically learned augmentation policy in the past two years. Specifically, recent work has empirically shown that the superior performance of the automated data augmentation methods stems from increasing the diversity of augmented data [4], [5]. However, two factors regarding the diversity of augmented data are still missing: 1) the explicit definition (and thus measurement) of diversity and 2) the quantifiable relationship between diversity and its regularization effects. To bridge this gap, we propose a diversity measure called Variance Diversity and theoretically show that the regularization effect of data augmentation is promised by Variance Diversity. We validate in experiments that the relative gain from automated data augmentation in test accuracy is highly correlated to Variance Diversity. An unsupervised sampling-based framework, DivAug, is designed to directly maximize Variance Diversity and hence strengthen the regularization effect. Without requiring a separate search process, the performance gain from DivAug is comparable with the state-of-the-art method with better efficiency. Moreover, under the semi-supervised setting, our framework can further improve the performance of semi-supervised learning algorithms compared to RandAugment, making it highly applicable to real-world problems, where labeled data is scarce. The code is available at https://github.com/warai-0toko/DivAug.

Outlier detection is an important data mining task with numerous applications such as intrusion detection, credit card fraud detection, and video surveillance. However, given a specific task with complex data, the process of building an effective deep learning based system for outlier detection still highly relies on human expertise and laboring trials. Moreover, while Neural Architecture Search (NAS) has shown its promise in discovering effective deep architectures in various domains, such as image classification, object detection and semantic segmentation, contemporary NAS methods are not suitable for outlier detection due to the lack of intrinsic search space and low sample efficiency. To bridge the gap, in this paper, we propose AutoOD, an automated outlier detection framework, which aims to search for an optimal neural network model within a predefined search space. Specifically, we introduce an experience replay mechanism based on self-imitation learning to improve the sample efficiency. Experimental results on various real-world benchmark datasets demonstrate that the deep model identified by AutoOD achieves the best performance, comparing with existing handcrafted models and traditional search methods.

Games are abstractions of the real world, where artificial agents learn to compete and cooperate with other agents. While significant achievements have been made in various perfect- and imperfect-information games, DouDizhu (a.k.a. Fighting the Landlord), a three-player card game, is still unsolved. DouDizhu is a very challenging domain with competition, collaboration, imperfect information, large state space, and particularly a massive set of possible actions where the legal actions vary significantly from turn to turn. Unfortunately, modern reinforcement learning algorithms mainly focus on simple and small action spaces, and not surprisingly, are shown not to make satisfactory progress in DouDizhu. In this work, we propose a conceptually simple yet effective DouDizhu AI system, namely DouZero, which enhances traditional Monte-Carlo methods with deep neural networks, action encoding, and parallel actors. Starting from scratch in a single server with four GPUs, DouZero outperformed all the existing DouDizhu AI programs in days of training and was ranked the first in the Botzone leaderboard among 344 AI agents. Through building DouZero, we show that classic Monte-Carlo methods can be made to deliver strong results in a hard domain with a complex action space. The code and an online demo are released at https://github.com/kwai/DouZero with the hope that this insight could motivate future work.

Back-propagation based visualizations have been proposed to interpret deep neural networks (DNNs), some of which produce interpretations with good visual quality. However, there exist doubts about whether these intuitive visualizations are related to network decisions. Recent studies have confirmed this suspicion by verifying that almost all these modified back-propagation visualizations are not faithful to the model's decision-making process. Besides, these visualizations produce vague "relative importance scores", among which low values can't guarantee to be independent of the final prediction. Hence, it's highly desirable to develop a novel back-propagation method that guarantees theoretical faithfulness and produces a quantitative attribution score with a clear understanding. To achieve the goal, we resort to mutual information theory to generate the interpretations, studying how much information of output is encoded in each input neuron. The basic idea is to learn a source signal by back-propagation such that the mutual information between input and output should be as much as possible preserved in the mutual information between input and the source signal. In addition, we propose a Mutual Information Preserving Inverse Network, termed MIP-IN, in which the parameters of each layer are recursively trained to learn how to invert. During the inversion, forward relu operation is adopted to adapt the general interpretations to the specific input. We then empirically demonstrate that the inverted source signal satisfies completeness and minimality property, which are crucial for a faithful interpretation. Furthermore, the empirical study validates the effectiveness of interpretations generated by MIP-IN.

Counterfactuals, serving as one of the emerging type of model interpretations, have recently received attention from both researchers and practitioners. Counterfactual explanations formalize the exploration of "what-if'' scenarios, and are an instance of example-based reasoning using a set of hypothetical data samples. Counterfactuals essentially show how the model decision alters with input perturbations. Existing methods for generating counterfactuals are mainly algorithm-based, which are time-inefficient and assume the same counterfactual universe for different queries. To address these limitations, we propose a Model-based Counterfactual Synthesizer (MCS) framework for interpreting machine learning models. We first analyze the model-based counterfactual process and construct a base synthesizer using a conditional generative adversarial net (CGAN). To better approximate the counterfactual universe for those rare queries, we novelly employ the umbrella sampling technique to conduct the MCS framework training. Besides, we also enhance the MCS framework by incorporating the causal dependence among attributes with model inductive bias, and validate its design correctness from the causality identification perspective. Experimental results on several datasets demonstrate the effectiveness as well as efficiency of our proposed MCS framework, and verify the advantages compared with other alternatives.

Mengnan Du, Varun Manjunatha, Rajiv Jain, Ruchi Deshpande, Franck Dernoncourt, Jiuxiang Gu, Tong Sun, Xia Hu. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Existing bias mitigation methods for DNN models primarily work on learning debiased encoders. This process not only requires a lot of instance-level annotations for sensitive attributes, it also does not guarantee that all fairness sensitive information has been removed from the encoder. To address these limitations, we explore the following research question: Can we reduce the discrimination of DNN models by only debiasing the classification head, even with biased representations as inputs? To this end, we propose a new mitigation technique, namely, Representation Neutralization for Fairness (RNF) that achieves fairness by debiasing only the task-specific classification head of DNN models. To this end, we leverage samples with the same ground-truth label but different sensitive attributes, and use their neutralized representations to train the classification head of the DNN model. The key idea of RNF is to discourage the classification head from capturing spurious correlation between fairness sensitive information in encoder representations with specific class labels. To address low-resource settings with no access to sensitive attribute annotations, we leverage a bias-amplified model to generate proxy annotations for sensitive attributes. Experimental results over several benchmark datasets demonstrate our RNF framework to effectively reduce discrimination of DNN models with minimal degradation in task-specific performance.

Graph neural networks (GNNs) integrate deep architectures and topological structure modeling in an effective way. However, the performance of existing GNNs would decrease significantly when they stack many layers, because of the over-smoothing issue. Node embeddings tend to converge to similar vectors when GNNs keep recursively aggregating the representations of neighbors. To enable deep GNNs, several methods have been explored recently. But they are developed from either techniques in convolutional neural networks or heuristic strategies. There is no generalizable and theoretical principle to guide the design of deep GNNs. To this end, we analyze the bottleneck of deep GNNs by leveraging the Dirichlet energy of node embeddings, and propose a generalizable principle to guide the training of deep GNNs. Based on it, a novel deep GNN framework -- Energetic Graph Neural Networks (EGNN) is designed. It could provide lower and upper constraints in terms of Dirichlet energy at each layer to avoid over-smoothing. Experimental results demonstrate that EGNN achieves state-of-the-art performance by using deep layers.

Recent methods in sequential recommendation focus on learning an overall embedding vector from a user's behavior sequence for the next-item recommendation. However, from empirical analysis, we discovered that a user's behavior sequence often contains multiple conceptually distinct items, while a unified embedding vector is primarily affected by one's most recent frequent actions. Thus, it may fail to infer the next preferred item if conceptually similar items are not dominant in recent interactions. To this end, an alternative solution is to represent each user with multiple embedding vectors encoding different aspects of the user's intentions. Nevertheless, recent work on multi-interest embedding usually considers a small number of concepts discovered via clustering, which may not be comparable to the large pool of item categories in real systems. It is a non-trivial task to effectively model a large number of diverse conceptual prototypes, as items are often not conceptually well clustered in fine granularity. Besides, an individual usually interacts with only a sparse set of concepts. In light of this, we propose a novel Sparse Interest NEtwork (SINE) for sequential recommendation. Our sparse-interest module can adaptively infer a sparse set of concepts for each user from the large concept pool and output multiple embeddings accordingly. Given multiple interest embeddings, we develop an interest aggregation module to actively predict the user's current intention and then use it to explicitly model multiple interests for next-item prediction. Empirical results on several public benchmark datasets and one large-scale industrial dataset demonstrate that SINE can achieve substantial improvement over state-of-the-art methods.

Image captioning has made substantial progress with huge supporting image collections sourced from the web. However, recent studies have pointed out that captioning datasets, such as COCO, contain gender bias found in web corpora. As a result, learning models could heavily rely on the learned priors and image context for gender identification, leading to incorrect or even offensive errors. To encourage models to learn correct gender features, we reorganize the COCO dataset and present two new splits COCO-GB V1 and V2 datasets where the train and test sets have different gender-context joint distribution. Models relying on contextual cues will suffer from huge gender prediction errors on the anti-stereotypical test data. Benchmarking experiments reveal that most captioning models learn gender bias, leading to high gender prediction errors, especially for women. To alleviate the unwanted bias, we propose a new Guided Attention Image Captioning model (GAIC) which provides self-guidance on visual attention to encourage the model to capture correct gender visual evidence. Experimental results validate that GAIC can significantly reduce gender prediction errors with a competitive caption quality. Our codes and the designed benchmark datasets are available at https://github.com/datamllab/Mitigating_Gender_Bias_In_Captioning_System.

Dimensionality reduction is considered as an important step for ensuring competitive performance in unsupervised learning such as anomaly detection. Non-negative matrix factorization (NMF) is a widely used method to accomplish this goal. But NMF do not have the provision to include the neighborhood structure information and, as a result, may fail to provide satisfactory performance in presence of nonlinear manifold structure. To address this shortcoming, we propose to consider the neighborhood structural similarity information within the NMF framework and do so by modeling the data through a minimum spanning tree. We label the resulting method as the neighborhood structure-assisted NMF. We further develop both offline and online algorithms for implementing the proposed method. Empirical comparisons using twenty benchmark data sets as well as an industrial data set extracted from a hydropower plant demonstrate the superiority of the neighborhood structure-assisted NMF. Looking closer into the formulation and properties of the proposed NMF method and comparing it with several NMF variants reveal that inclusion of the MST-based neighborhood structure plays a key role in attaining the enhanced performance in anomaly detection.

Interpersonal ties describe the intensity of information and activity interactions among individuals. It plays a critical role in social network analysis and sociological studies. Existing efforts focus on leveraging individuals' non-structural characteristics to measure tie strength. With the booming of online social networks (OSNs), it has become difficult to process and measure all the non-structural data. We study the tie strength measuring from the network topological aspect. However, it remains a nontrivial task due to the controversial comprehensions of its definition and the large volume of OSNs. To tackle the challenges, we develop a scalable measuring framework - IETSM. From the network view, we formally define the tie strength of an edge as the inverse of its impact on the similarity between its two nodes' influences in information diffusion. To measure this impact, IETSM constructs a node's influence as the embedding learned from its neighborhoods inductively. It estimates the tie strength of an edge through its impact on its nodes' influences brought by deleting it. The learned tie strength scores could, in turn, facilitate the node representation learning, and we update them iteratively. Experiments on real-world datasets demonstrate the effectiveness and efficiency of IETSM.

With advancements of deep learning techniques, it is now possible to generate super-realistic images and videos, i.e., deepfakes. These deepfakes could reach mass audience and result in adverse impacts on our society. Although lots of efforts have been devoted to detect deepfakes, their performance drops significantly on previously unseen but related manipulations and the detection generalization capability remains a problem. Motivated by the fine-grained nature and spatial locality characteristics of deepfakes, we propose Locality-Aware AutoEncoder (LAE) to bridge the generalization gap. In the training process, we use a pixel-wise mask to regularize local interpretation of LAE to enforce the model to learn intrinsic representation from the forgery region, instead of capturing artifacts in the training set and learning superficial correlations to perform detection. We further propose an active learning framework to select the challenging candidates for labeling, which requires human masks for less than 3% of the training data, dramatically reducing the annotation efforts to regularize interpretations. Experimental results on three deepfake detection tasks indicate that LAE could focus on the forgery regions to make decisions. The analysis further shows that LAE outperforms the state-of-the-arts by 6.52%, 12.03%, and 3.08% respectively on three deepfake detection tasks in terms of generalization accuracy on previously unseen manipulations.

Recommender systems play a fundamental role in web applications in filtering massive information and matching user interests. While many efforts have been devoted to developing more effective models in various scenarios, the exploration on the explainability of recommender systems is running behind. Explanations could help improve user experience and discover system defects. In this paper, after formally introducing the elements that are related to model explainability, we propose a novel explainable recommendation model through improving the transparency of the representation learning process. Specifically, to overcome the representation entangling problem in traditional models, we revise traditional graph convolution to discriminate information from different layers. Also, each representation vector is factorized into several segments, where each segment relates to one semantic aspect in data. Different from previous work, in our model, factor discovery and representation learning are simultaneously conducted, and we are able to handle extra attribute information and knowledge. In this way, the proposed model can learn interpretable and meaningful representations for users and items. Unlike traditional methods that need to make a trade-off between explainability and effectiveness, the performance of our proposed explainable model is not negatively affected after considering explainability. Finally, comprehensive experiments are conducted to validate the performance of our model as well as explanation faithfulness.

High false-positive rate is a long-standing challenge for anomaly detection algorithms, especially in high-stake applications. To identify the true anomalies, in practice, analysts or domain experts will be employed to investigate the top instances one by one in a ranked list of anomalies identified by an anomaly detection system. This verification procedure generates informative labels that can be leveraged to re-rank the anomalies so as to help the analyst to discover more true anomalies given a time budget. Some re-ranking strategies have been proposed to approximate the above sequential decision process. Specifically, existing strategies have been focused on making the top instances more likely to be anomalous based on the feedback. Then they greedily select the top-1 instance for query. However, these greedy strategies could be sub-optimal since some low-ranked instances could be more helpful in the long-term. Motivated by this, in this work, we study whether modeling long-term performance can benefit active anomaly detection. This is a challenging task because it is unclear how long-term performance could be quantified. In addition, the query selection has a huge decision space, which is difficult to model. To address these challenges, we propose Active Anomaly Detection with Meta-Policy (Meta-AAD), a novel framework that learns a meta-policy for query selection. Specifically, Meta-AAD leverages deep reinforcement learning to train the meta-policy to select the most proper instance to explicitly optimize the number of discovered anomalies throughout the querying process. Meta-AAD is easy to deploy since a trained meta-policy can be directly applied to any new datasets without further tuning. Extensive experiments on 24 benchmark datasets demonstrate that Meta-AAD significantly outperforms the state-of-the-art re-ranking strategies and the unsupervised baseline. The empirical analysis shows that the trained meta-policy is transferable and inherently achieves a balance between long-term and short-term rewards.

Policy distillation, which transfers a teacher policy to a student policy has achieved great success in challenging tasks of deep reinforcement learning. This teacher-student framework requires a well-trained teacher model which is computationally expensive. Moreover, the performance of the student model could be limited by the teacher model if the teacher model is not optimal. In the light of collaborative learning, we study the feasibility of involving joint intellectual efforts from diverse perspectives of student models. In this work, we introduce dual policy distillation (DPD), a student-student framework in which two learners operate on the same environment to explore different perspectives of the environment and extract knowledge from each other to enhance their learning. The key challenge in developing this dual learning framework is to identify the beneficial knowledge from the peer learner for contemporary learning-based reinforcement learning algorithms, since it is unclear whether the knowledge distilled from an imperfect and noisy peer learner would be helpful. To address the challenge, we theoretically justify that distilling knowledge from a peer learner will lead to policy improvement and propose a disadvantageous distillation strategy based on the theoretical results. The conducted experiments on several continuous control tasks show that the proposed framework achieves superior performance with a learning-based agent and function approximation without the use of expensive teacher models.

We present RLCard, a Python platform for reinforcement learning research and development in card games. RLCard supports various card environments and several baseline algorithms with unified easy-to-use interfaces, aiming at bridging reinforcement learning and imperfect information games. The platform provides flexible configurations of state representation, action encoding, and reward design. RLCard also supports visualizations for algorithm debugging. In this demo, we showcase two representative environments and their visualization results. We conclude this demo with challenges and research opportunities brought by RLCard. A video is available on YouTube.

The classification of graph-structured data has be-come increasingly crucial in many disciplines. It has been observed that the implicit or explicit hierarchical community structures preserved in real-world graphs could be useful for downstream classification applications. A straightforward way to leverage the hierarchical structure is to make use the pooling algorithms to cluster nodes into fixed groups, and shrink the input graph layer by layer to learn the pooled graphs.However, the pool shrinking discards the graph details to make it hard to distinguish two non-isomorphic graphs, and the fixed clustering ignores the inherent multiple characteristics of nodes. To compensate the shrinking loss and learn the various nodes’ characteristics, we propose the multi-channel graph neural networks (MuchGNN). Motivated by the underlying mechanisms developed in convolutional neural networks, we define the tailored graph convolutions to learn a series of graph channels at each layer, and shrink the graphs hierarchically to en-code the pooled structures. Experimental results on real-world datasets demonstrate the superiority of MuchGNN over the state-of-the-art methods.

Graph data are pervasive in many real-world applications. Recently, increasing attention has been paid on graph neural networks (GNNs), which aim to model the local graph structures and capture the hierarchical patterns by aggregating the information from neighbors with stackable network modules. Motivated by the observation that different nodes often require different iterations of aggregation to fully capture the structural information, in this paper, we propose to explicitly sample diverse iterations of aggregation for different nodes to boost the performance of GNNs. It is a challenging task to develop an effective aggregation strategy for each node, given complex graphs and sparse features. Moreover, it is not straightforward to derive an efficient algorithm since we need to feed the sampled nodes into different number of network layers. To address the above challenges, we propose Policy-GNN, a meta-policy framework that models the sampling procedure and message passing of GNNs into a combined learning process. Specifically, Policy-GNN uses a meta-policy to adaptively determine the number of aggregations for each node. The meta-policy is trained with deep reinforcement learning~(RL) by exploiting the feedback from the model. We further introduce parameter sharing and a buffer mechanism to boost the training efficiency. Experimental results on three real-world benchmark datasets suggest that Policy-GNN significantly outperforms the state-of-the-art alternatives, showing the promise in aggregation optimization for GNNs.

Click-Through Rate (CTR) prediction is one of the most important machine learning tasks in recommender systems, driving personalized experience for billions of consumers. Neural architecture search (NAS), as an emerging field, has demonstrated its capabilities in discovering powerful neural network architectures, which motivates us to explore its potential for CTR predictions. Due to 1) diverse unstructured feature interactions, 2) heterogeneous feature space, and 3) high data volume and intrinsic data randomness, it is challenging to construct, search, and compare different architectures effectively for recommendation models. To address these challenges, we propose an automated interaction architecture discovering framework for CTR prediction named AutoCTR. Via modularizing simple yet representative interactions as virtual building blocks and wiring them into a space of direct acyclic graphs, AutoCTR performs evolutionary architecture exploration with learning-to-rank guidance at the architecture level and achieves acceleration using low-fidelity model. Empirical analysis demonstrates the effectiveness of AutoCTR on different datasets comparing to human-crafted architectures. The discovered architecture also enjoys generalizability and transferability among different datasets.

With the widespread use of deep neural networks (DNNs) in high-stake applications, the security problem of the DNN models has received extensive attention. In this paper, we investigate a specific security problem called trojan attack, which aims to attack deployed DNN systems relying on the hidden trigger patterns inserted by malicious hackers. We propose a training-free attack approach which is different from previous work, in which trojaned behaviors are injected by retraining model on a poisoned dataset. Specifically, we do not change parameters in the original model but insert a tiny trojan module (TrojanNet) into the target model. The infected model with a malicious trojan can misclassify inputs into a target label when the inputs are stamped with the special trigger. The proposed TrojanNet has several nice properties including (1) it activates by tiny trigger patterns and keeps silent for other signals, (2) it is model-agnostic and could be injected into most DNNs, dramatically expanding its attack scenarios, and (3) the training-free mechanism saves massive training efforts comparing to conventional trojan attack methods. The experimental results show that TrojanNet can inject the trojan into all labels simultaneously (all-label trojan attack) and achieves 100% attack success rate without affecting model accuracy on original tasks. Experimental analysis further demonstrates that state-of-the-art trojan detection algorithms fail to detect TrojanNet attack. The code is available at https://github.com/trx14/TrojanNet.

Graphs neural networks (GNNs) learn node features by aggregating and combining neighbor information, which have achieved promising performance on many graph tasks. However, GNNs are mostly treated as black-boxes and lack human intelligible explanations. Thus, they cannot be fully trusted and used in certain application domains if GNN models cannot be explained. In this work, we propose a novel approach, known as XGNN, to interpret GNNs at the model-level. Our approach can provide high-level insights and generic understanding of how GNNs work. In particular, we propose to explain GNNs by training a graph generator so that the generated graph patterns maximize a certain prediction of the model. We formulate the graph generation as a reinforcement learning task, where for each step, the graph generator predicts how to add an edge into the current graph. The graph generator is trained via a policy gradient method based on information from the trained GNNs. In addition, we incorporate several graph rules to encourage the generated graphs to be valid. Experimental results on both synthetic and real-world datasets show that our proposed methods help understand and verify the trained GNNs. Furthermore, our experimental results indicate that the generated graphs can provide guidance on how to improve the trained GNNs.

Detecting statistical interactions between input features is a crucial and challenging task. Recent advances demonstrate that it is possible to extract learned interactions from trained neural networks. It has also been observed that, in neural networks, any interacting features must follow a strongly weighted connection to common hidden units. Motivated by the observation, in this paper, we propose to investigate the interaction detection problem from a novel topological perspective by analyzing the connectivity in neural networks. Specially, we propose a new measure for quantifying interaction strength, based upon the well-received theory of persistent homology. Based on this measure, a Persistence Interaction Dection (PID) algorithm is developed to efficiently detect interactions. Our proposed algorithm is evaluated across a number of interaction detection tasks on several synthetic and real-world datasets with different hyperparameters. Experimental results validate that the PID algorithm outperforms the state-of-the-art baselines.

Graph neural networks (GNNs), which learn the representation of a node by aggregating its neighbors, have become an effective computational tool in downstream applications. Over-smoothing is one of the key issues which limit the performance of GNNs as the number of layers increases. It is because the stacked aggregators would make node representations converge to indistinguishable vectors. Several attempts have been made to tackle the issue by bringing linked node pairs close and unlinked pairs distinct. However, they often ignore the intrinsic community structures and would result in sub-optimal performance. The representations of nodes within the same community/class need be similar to facilitate the classification, while different classes are expected to be separated in embedding space. To bridge the gap, we introduce two over-smoothing metrics and a novel technique, i.e., differentiable group normalization (DGN). It normalizes nodes within the same group independently to increase their smoothness, and separates node distributions among different groups to significantly alleviate the over-smoothing issue. Experiments on real-world datasets demonstrate that DGN makes GNN models more robust to over-smoothing and achieves better performance with deeper GNNs.

Realistic recommender systems are often required to adapt to ever-changing data and tasks or to explore different models systematically. To address the need, we present AutoRec 1 2, an open-source automated machine learning (AutoML) platform extended from the TensorFlow [3] ecosystem and, to our knowledge, the first framework to leverage AutoML for model search and hyperparameter tuning in deep recommendation models. AutoRec also supports a highly flexible pipeline that accommodates both sparse and dense inputs, rating prediction and click-through rate (CTR) prediction tasks, and an array of recommendation models. Lastly, AutoRec provides a simple, user-friendly API. Experiments conducted on the benchmark datasets reveal AutoRec is reliable and can identify models which resemble the best model without prior knowledge.

Deep neural network (DNN) has become an effective computational tool because of its superior performance in practice. However, the generalization of DNN still largely depends on the training data, no matter in quantity or quality. In this paper, we propose a knowledge instillation framework, named NeuKI, for feed-forward DNN, aiming to enhance learning performance with the aid of knowledge. This task is particularly challenging due to the complicated nature of knowledge and numerous variants of DNN architectures. To bridge the gap, we construct a separate knowledge-DNN faithfully encoding the instilled knowledge for joint training. The core idea is to regularize the training of target-DNN with the constructed knowledge-DNN, so that the instilled knowledge can guide the model training. The proposed NeuKI is demonstrated to be applicable to both knowledge rules and constraints, where rules are encoded by structure and constraints are handled by loss. Experiments are conducted on several real-world datasets from different domains, and the results demonstrate the effectiveness of NeuKI in improving learning performance, as well as relevant data efficiency and model interpretability.

Recommender systems in industry generally include two stages: recall and ranking. Recall refers to efficiently identify hundreds of candidate items that user may interest in from a large volume of item corpus, while the latter aims to output a precise ranking list using complex ranking models. Recently, graph representation learning has attracted much attention in supporting high quality candidate search at scale. Despite its effectiveness in learning embedding vectors for objects in the user-item interaction network, the computational costs to infer users’ preferences in continuous embedding space are tremendous. In this work, we investigate the problem of hashing with graph neural networks (GNNs) for high quality retrieval, and propose a simple yet effective discrete representation learning framework to jointly learn continuous and discrete codes. Specifically, a deep hashing with GNNs (HashGNN) is presented, which consists of two components, a GNN encoder for learning node representations, and a hash layer for encoding representations to hash codes. The whole architecture is trained end-to-end by jointly optimizing two losses, i.e., reconstruction loss from reconstructing observed links, and ranking loss from preserving the relative ordering of hash codes. A novel discrete optimization strategy based on straight through estimator (STE) with guidance is proposed. The principal idea is to avoid gradient magnification in back-propagation of STE with continuous embedding guidance, in which we begin from learning an easier network that mimic the continuous embedding and let it evolve during the training until it finally goes back to STE. Comprehensive experiments over three publicly available and one real-world Alibaba company datasets demonstrate that our model not only can achieve comparable performance compared with its continuous counterpart but also runs multiple times faster during inference.

Attributed graphs, which contain rich contextual features beyond just network structure, are ubiquitous and have been observed to benefit various network analytics applications. Graph structure optimization, aiming to find the optimal graphs in terms of some specific measures, has become an effective computational tool in complex network analysis. However, traditional model-free methods suffer from the expensive computational cost of evaluating graphs; existing vectorial Bayesian optimization methods cannot be directly applied to attributed graphs and have the scalability issue due to the use of Gaussian processes (GPs). To bridge the gap, in this paper, we propose a novel scalable Deep Graph Bayesian Optimization (DGBO) method on attributed graphs. The proposed DGBO prevents the cubical complexity of the GPs by adopting a deep graph neural network to surrogate black-box functions, and can scale linearly with the number of observations. Intensive experiments are conducted on both artificial and real-world problems, including molecular discovery and urban road network design, and demonstrate the effectiveness of the DGBO compared with the state-of-the-art.

Feature embedding aims to learn a low-dimensional vector representation for each instance to preserve the information in its features. These representations can benefit various offthe-shelf learning algorithms. While embedding models for a single type of features have been well-studied, real-world instances often contain multiple types of correlated features or even information within a different modality such as networks. Existing studies such as multiview learning show that it is promising to learn unified vector representations from all sources. However, high computational costs of incorporating heterogeneous information limit the applications of existing algorithms. The number of instances and dimensions of features in practice are often large. To bridge the gap, we propose a scalable framework FeatWalk, which can model and incorporate instance similarities in terms of different types of features into a unified embedding representation. To enable the scalability, FeatWalk does not directly calculate any similarity measure, but provides an alternative way to simulate the similarity-based random walks among instances to extract the local instance proximity and preserve it in a set of instance index sequences. These sequences are homogeneous with each other. A scalable word embedding algorithm is applied to them to learn a joint embedding representation of instances. Experiments on four real-world datasets demonstrate the efficiency and effectiveness of FeatWalk.

Interpreting deep neural networks is of great importance to understand and verify deep models for natural language processing (NLP) tasks. However, most existing approaches only focus on improving the performance of models but ignore their interpretability. In this work, we propose an approach to investigate the meaning of hidden neurons of the convolutional neural network (CNN) models. We first employ saliency map and optimization techniques to approximate the detected information of hidden neurons from input sentences. Then we develop regularization terms and explore words in vocabulary to interpret such detected information. Experimental results demonstrate that our approach can identify meaningful and reasonable interpretations for hidden spatial locations. Additionally, we show that our approach can describe the decision procedure of deep NLP models.

Recent explainability related studies have shown that state-of-the-art DNNs do not always adopt correct evidences to make decisions. It not only hampers their generalization but also makes them less likely to be trusted by end-users. In pursuit of developing more credible DNNs, in this paper we propose CREX, which encourages DNN models to focus more on evidences that actually matter for the task at hand, and to avoid overfitting to data-dependent bias and artifacts. Specifically, CREX regularizes the training process of DNNs with rationales, i.e., a subset of features highlighted by domain experts as justifications for predictions, to enforce DNNs to generate local explanations that conform with expert rationales. Even when rationales are not available, CREX still could be useful by requiring the generated explanations to be sparse. Experimental results on two text classification datasets demonstrate the increased credibility of DNNs trained with CREX. Comprehensive analysis further shows that while CREX does not always improve prediction accuracy on the held-out test set, it significantly increases DNN accuracy on new and previously unseen data beyond test set, highlighting the advantage of the increased credibility.

Experience replay enables reinforcement learning agents to memorize and reuse past experiences, just as humans replay memories for the situation at hand. Contemporary off-policy algorithms either replay past experiences uniformly or utilize a rule-based replay strategy, which may be sub-optimal. In this work, we consider learning a replay policy to optimize the cumulative reward. Replay learning is challenging because the replay memory is noisy and large, and the cumulative reward is unstable. To address these issues, we propose a novel experience replay optimization (ERO) framework which alternately updates two policies: the agent policy, and the replay policy. The agent is updated to maximize the cumulative reward based on the replayed data, while the replay policy is updated to provide the agent with the most useful experiences. The conducted experiments on various continuous control tasks demonstrate the effectiveness of ERO, empirically showing promise in experience replay learning to improve the performance of off-policy reinforcement learning algorithms.

Anomaly detection is a fundamental problem in data mining field with many real-world applications. A vast majority of existing anomaly detection methods predominately focused on data collected from a single source. In real-world applications, instances often have multiple types of features, such as images (ID photos, finger prints) and texts (bank transaction histories, user online social media posts), resulting in the so-called multi-modal data. In this paper, we focus on identifying anomalies whose patterns are disparate across different modalities, i.e., cross-modal anomalies. Some of the data instances within a multi-modal context are often not anomalous when they are viewed separately in each individual modality, but contains inconsistent patterns when multiple sources are jointly considered. The existence of multi-modal data in many real-world scenarios brings both opportunities and challenges to the canonical task of anomaly detection. On the one hand, in multimodal data, information of different modalities may complement each other in improving the detection performance. On the other hand, complicated distributions across different modalities call for a principled framework to characterize their inherent and complex correlations, which is often difficult to capture with conventional linear models. To this end, we propose a novel deep structured anomaly detection framework to identify the cross-modal anomalies embedded in the data. Experiments on real-world datasets demonstrate the effectiveness of the proposed framework comparing with the state-of-the-art.

Random walks are widely adopted in various network analysis tasks ranging from network embedding to label propagation. It could capture and convert geometric structures into structured sequences while alleviating the issues of sparsity and curse of dimensionality. Though random walks on plain networks have been intensively studied, in real-world systems, nodes are often not pure vertices, but own different characteristics, described by the rich set of data associated with them. These node attributes contain plentiful information that often complements the network, and bring opportunities to the random-walk-based analysis. However, it is unclear how random walks could be developed for attributed networks towards an effective joint information extraction. Node attributes make the node interactions more complicated and are heterogeneous with respect to topological structures.

Neural architecture search (NAS) has been proposed to automatically tune deep neural networks, but existing search algorithms, e.g., NASNet, PNAS, usually suffer from expensive computational cost. Network morphism, which keeps the functionality of a neural network while changing its neural architecture, could be helpful for NAS by enabling more efficient training during the search. In this paper, we propose a novel framework enabling Bayesian optimization to guide the network morphism for efficient neural architecture search. The framework develops a neural network kernel and a tree-structured acquisition function optimization algorithm to efficiently explores the search space. Extensive experiments on real-world benchmark datasets have been done to demonstrate the superior performance of the developed framework over the state-of-the-art methods. Moreover, we build an open-source AutoML system based on our method, namely Auto-Keras. The code and documentation are available at https://autokeras.com. The system runs in parallel on CPU and GPU, with an adaptive search strategy for different GPU memory limits.

Networks have been widely used as the data structure for abstracting real-world systems as well as organizing the relations among entities. Network embedding models are powerful tools in mapping nodes in a network into continuous vector-space representations in order to facilitate subsequent tasks such as classification and link prediction. Existing network embedding models comprehensively integrate all information of each node, such as links and attributes, towards a single embedding vector to represent the node's general role in the network. However, a real-world entity could be multifaceted, where it connects to different neighborhoods due to different motives or self-characteristics that are not necessarily correlated. For example, in a movie recommender system, a user may love comedies or horror movies simultaneously, but it is not likely that these two types of movies are mutually close in the embedding space, nor the user embedding vector could be sufficiently close to them at the same time. In this paper, we propose a polysemous embedding approach for modeling multiple facets of nodes, as motivated by the phenomenon of word polysemy in language modeling. Each facet of a node is mapped to an embedding vector, while we also maintain an association degree between each pair of node and facet. The proposed method is adaptive to various existing embedding models, without significantly complicating the optimization process. We also discuss how to engage embedding vectors of different facets for inference tasks including classification and link prediction. Experiments on real-world datasets help comprehensively evaluate the performance of the proposed method.

We focus on the problem of streaming recommender system and explore novel collaborative filtering algorithms to handle the data dynamicity and complexity in a streaming manner. Although deep neural networks have demonstrated the effectiveness of recommendation tasks, it is lack of explorations on integrating probabilistic models and deep architectures under streaming recommendation settings. Conjoining the complementary advantages of probabilistic models and deep neural networks could enhance both model effectiveness and the understanding of inference uncertainties. To bridge the gap, in this paper, we propose a Coupled Variational Recurrent Collaborative Filtering (CVRCF) framework based on the idea of Deep Bayesian Learning to handle the streaming recommendation problem. The framework jointly combines stochastic processes and deep factorization models under a Bayesian paradigm to model the generation and evolution of users' preferences and items' popularities. To ensure efficient optimization and streaming update, we further propose a sequential variational inference algorithm based on a cross variational recurrent neural network structure. Experimental results on three benchmark datasets demonstrate that the proposed framework performs favorably against the state-of-the-art methods in terms of both temporal dependency modeling and predictive accuracy. The learned latent variables also provide visualized interpretations for the evolution of temporal dynamics.

Representation learning models map data instances into a low-dimensional vector space, thus facilitating the deployment of subsequent models such as classification and clustering models, or the implementation of downstream applications such as recommendation and anomaly detection. However, the outcome of representation learning is difficult to be directly understood by users, since each dimension of the latent space may not have any specific meaning. Understanding representation learning could be beneficial to many applications. For example, in recommender systems, knowing why a user instance is mapped to a certain position in the latent space may unveil the user's interests and profile. In this paper, we propose an interpretation framework to understand and describe how representation vectors distribute in the latent space. Specifically, we design a coding scheme to transform representation instances into spatial codes to indicate their locations in the latent space. Following that, a multimodal autoencoder is built for generating the description of a representation instance given its spatial codes. The coding scheme enables indication of position with different granularity. The incorporation of autoencoder makes the framework capable of dealing with different types of data. Several metrics are designed to evaluate interpretation results. Experiments under various application scenarios and different representation learning models are conducted to demonstrate the flexibility and effectiveness of the proposed framework.

RNN models have achieved the state-of-the-art performance in a wide range of text mining tasks. However, these models are often regarded as black-boxes and are criticized due to the lack of interpretability. In this paper, we enhance the interpretability of RNNs by providing interpretable rationales for RNN predictions. Nevertheless, interpreting RNNs is a challenging problem. Firstly, unlike existing methods that rely on local approximation, we aim to provide rationales that are more faithful to the decision making process of RNN models. Secondly, a flexible interpretation method should be able to assign contribution scores to text segments of varying lengths, instead of only to individual words. To tackle these challenges, we propose a novel attribution method, called REAT, to provide interpretations to RNN predictions. REAT decomposes the final prediction of a RNN into additive contribution of each word in the input text. This additive decomposition enables REAT to further obtain phrase-level attribution scores. In addition, REAT is generally applicable to various RNN architectures, including GRU, LSTM and their bidirectional versions. Experimental results demonstrate the faithfulness and interpretability of the proposed attribution method. Comprehensive analysis shows that our attribution method could unveil the useful linguistic knowledge captured by RNNs. Some analysis further demonstrates our method could be utilized as a debugging tool to examine the vulnerability and failure reasons of RNNs, which may lead to several promising future directions to promote generalization ability of RNNs.

In this demo paper, we present the XFake system, an explainable fake news detector that assists end-users to identify news credibility. To effectively detect and interpret the fakeness of news items, we jointly consider both attributes (e.g., speaker) and statements. Specifically, MIMIC, ATTN and PERT frameworks are designed, where MIMIC is built for attribute analysis, ATTN is for statement semantic analysis and PERT is for statement linguistic analysis. Beyond the explanations extracted from the designed frameworks, relevant supporting examples as well as visualization are further provided to facilitate the interpretation. Our implemented system is demonstrated on a real-world dataset crawled from PolitiFact1, where thousands of verified political news have been collected.

MOTIVATION: Cellular function is closely related to the localizations of its sub-structures. It is, however, challenging to experimentally label all sub-cellular structures simultaneously in the same cell. This raises the need of building a computational model to learn the relationships among these sub-cellular structures and use reference structures to infer the localizations of other structures. RESULTS: We formulate such a task as a conditional image generation problem and propose to use conditional generative adversarial networks for tackling it. We employ an encoder-decoder network as the generator and propose to use skip connections between the encoder and decoder to provide spatial information to the decoder. To incorporate the conditional information in a variety of different ways, we develop three different types of skip connections, known as the self-gated connection, encoder-gated connection and label-gated connection. The proposed skip connections are built based on the conditional information using gating mechanisms. By learning a gating function, the network is able to control what information should be passed through the skip connections from the encoder to the decoder. Since the gate parameters are also learned automatically, we expect that only useful spatial information is transmitted to the decoder to help image generation. We perform both qualitative and quantitative evaluations to assess the effectiveness of our proposed approaches. Experimental results show that our cGAN-based approaches have the ability to generate the desired sub-cellular structures correctly. Our results also demonstrate that the proposed approaches outperform the existing approach based on adversarial auto-encoders, and the new skip connections lead to improved performance. In addition, the localizations of generated sub-cellular structures by our approaches are consistent with observations in biological experiments. AVAILABILITY AND IMPLEMENTATION: The source code and more results are available at https://github.com/divelab/cgan/.

Tensor completion is a problem of filling the missing or unobserved entries of partially observed tensors. Due to the multidimensional character of tensors in describing complex datasets, tensor completion algorithms and their applications have received wide attention and achievement in areas like data mining, computer vision, signal processing, and neuroscience. In this survey, we provide a modern overview of recent advances in tensor completion algorithms from the perspective of big data analytics characterized by diverse variety, large volume, and high velocity. We characterize these advances from the following four perspectives: general tensor completion algorithms, tensor completion with auxiliary information (variety), scalable tensor completion algorithms (volume), and dynamic tensor completion algorithms (velocity). Further, we identify several tensor completion applications on real-world data-driven problems and present some common experimental frameworks popularized in the literature along with several available software repositories. Our goal is to summarize these popular methods and introduce them to researchers and practitioners for promoting future research and applications. We conclude with a discussion of key challenges and promising research directions in this community for future exploration.

Link prediction in social networks is to infer the new links likely to be formed next or to reconstruct the links that are currently missing. Other than the pure topological network structures, social networks are often associated with rich information of social activities of users, such as tweeting, retweeting, and replying. Social theories such as social influence indicate that social activities could have potential impacts on the neighbors, and links in social media could be the results of the social influence among users. It motivates us to learn and model social influence among users to tackle the link prediction problem. However, this is a non-trivial task since it is challenging to model heterogeneous social activities. Traditional methods often define universal metrics of social influence for all users, but even for the same activity of a user, the influence towards different neighbors might not be the same. It motivates a personalized learning schema. In information theory, if a time-series signal influences another, then the uncertainty in the latter one will be reduced, given the distribution of the former one. Thus, we are motivated to learn social influence based on the timestamps of social activities. Given the timestamps of each user, we use entropy to measure the reduction of uncertainty of his/her neighbors. The learned social influence is then incorporated into a graph based link prediction model to perform joint learning. Through comprehensive experiments, we demonstrate that the proposed framework can perform better than the state-of-the-art methods on different real-world networks.

Tensor-based methods have shown promise in improving upon traditional matrix factorization methods for recommender systems. But tensors may achieve improved recommendation quality while worsening the fairness of the recommendations. Hence, we propose a novel fairness-aware tensor recommendation framework that is designed to maintain quality while dramatically improving fairness. Four key aspects of the proposed framework are: (i) a new sensitive latent factor matrix for isolating sensitive features; (ii) a sensitive information regularizer that extracts sensitive information which can taint other latent factors; (iii) an effective algorithm to solve the proposed optimization model; and (iv) extension to multi-feature and multi-category cases which previous efforts have not addressed. Extensive experiments on real-world and synthetic datasets show that the framework enhances recommendation fairness while preserving recommendation quality in comparison with state-of-the-art alternatives.

How can we efficiently recover missing values for very large-scale real-world datasets that are multi-dimensional even when the auxiliary information is regularized at certain mode? Tensor completion is a useful tool to recover a low-rank tensor that best approximates partially observed data and further predicts the unobserved data by this low-rank tensor, which has been successfully used for many applications such as location-based recommender systems, link prediction, targeted advertising, social media search, and event detection. Due to the curse of dimensionality, existing algorithms for tensor completion that integrate auxiliary information do not scale for tensors with billions of elements. In this paper, we propose DisTenC, a new distributed large-scale tensor completion algorithm that can be distributed on Spark. Our key insights are to (i) efficiently handle trace-based regularization terms; (ii) update factor matrices with caching; and (iii) optimize the update of the new tensor via residuals. In this way, we can tackle the high computational costs of traditional approaches and minimize intermediate data, leading to order-of-magnitude improvements in tensor completion. Experimental results demonstrate that DisTenC is capable of handling up to 10~1000X larger tensors than existing methods with much faster convergence rate, shows better linearity on machine scalability, and achieves up to an average improvement of 23.5% in accuracy in applications.

Adversarial examples are delicately perturbed inputs, which aim to mislead machine learning models towards incorrect outputs. While existing work focuses on generating adversarial perturbations in multiclass classification problems, many real-world applications fall into the multi-label setting, in which one instance could be associated with more than one label. To analyze the vulnerability and robustness of multi-label learning models, we investigate the generation of multi-label adversarial perturbations. This is a challenging task due to the uncertain number of positive labels associated with one instance, and the fact that multiple labels are usually not mutually exclusive with each other. To bridge the gap, in this paper, we propose a general attacking framework targeting multi-label classification problem and conduct a premier analysis on the perturbations for deep neural networks. Leveraging the ranking relationships among labels, we further design a ranking-based framework to attack multi-label ranking algorithms. Experiments on two different datasets demonstrate the effectiveness of the proposed frameworks and provide insights of the vulnerability of multi-label deep models under diverse targeted attacks.

Despite the wide application in recent years, most recommender systems are not capable of providing interpretations together with recommendation results, which impedes both deployers and customers from understanding or trusting the results. Recent advances in recommendation models, such as deep learning models, usually involve extracting latent representations of users and items. However, the representation space is not directly comprehensible since each dimension usually does not have any specific meaning. In addition, recommender systems incorporate various sources of information, such as user behaviors, item information, and other side content information. Properly organizing different types of information, as well as effectively selecting important information for interpretation, is challenging and has not been fully tackled by conventional interpretation methods. In this paper, we propose a post-hoc method called Sorted Explanation Paths (SEP) to interpret recommendation results. Specifically, we first build a unified heterogeneous information network to incorporate multiple types of objects and relations based on representations from the recommender system and information from the dataset. Then, we search for explanation paths between given recommendation pairs, and use the set of simple paths to construct semantic explanations. Next, three heuristic metrics, i.e., credibility, readability and diversity, are designed to measure the validity of each explanation path, and to sort all the paths comprehensively. The top-ranked explanation paths are selected as the final interpretation. After that, practical issues on computation and efficiency of the proposed SEP method are also handled by corresponding approaches. Finally, we conduct experiments on three real-world benchmark datasets, and demonstrate the applicability and effectiveness of the proposed SEP method.

While outlier detection has been intensively studied in many applications, interpretation is becoming increasingly important to help people trust and evaluate the developed detection models through providing intrinsic reasons why the given outliers are identified. It is a nontrivial task for interpreting the abnormality of outliers due to the distinct characteristics of different detection models, complicated structures of data in certain applications, and imbalanced distribution of outliers and normal instances. In addition, contexts where outliers locate, as well as the relation between outliers and the contexts, are usually overlooked in existing interpretation frameworks. To tackle the issues, in this paper, we propose a Contextual Outlier INterpretation (COIN) framework to explain the abnormality of outliers spotted by detectors. The interpretability of an outlier is achieved through three aspects, i.e., outlierness score, attributes that contribute to the abnormality, and contextual description of its neighborhoods. Experimental results on various types of datasets demonstrate the flexibility and effectiveness of the proposed framework.

User-generated social media data are exploding and of high demand in public and private sectors. The disclosure of intact social media data exacerbates the threats to user privacy. In this paper, we first identify a text-based user-linkage attack on current data outsourcing practices, in which the real users in an anonymized dataset can be pinpointed based on the users' unprotected text data. Then we propose a framework for differentially privacy-preserving social media data outsourcing for the first time in literature. Within our framework, social media data service providers can outsource perturbed datasets to provide users differential privacy while offering high data utility to social media data consumers. Our differential privacy mechanism is based on a novel notion of E - text indistinguishability, which we propose to thwart the text-based user-linkage attack. Extensive experiments on real-world and synthetic datasets confirm that our framework can enable high-level differential privacy protection and also high data utility.

While deep neural networks (DNN) have become an effective computational tool, the prediction results are often criticized by the lack of interpretability, which is essential in many real-world applications such as health informatics. Existing attempts based on local interpretations aim to identify relevant features contributing the most to the prediction of DNN by monitoring the neighborhood of a given input. They usually simply ignore the intermediate layers of the DNN that might contain rich information for interpretation. To bridge the gap, in this paper, we propose to investigate a guided feature inversion framework for taking advantage of the deep architectures towards effective interpretation. The proposed framework not only determines the contribution of each feature in the input but also provides insights into the decision-making process of DNN models. By further interacting with the neuron of the target category at the output layer of the DNN, we enforce the interpretation result to be class-discriminative. We apply the proposed interpretation model to different CNN architectures to provide explanations for image data and conduct extensive experiments on ImageNet and PASCAL VOC07 datasets. The interpretation results demonstrate the effectiveness of our proposed framework in providing class-discriminative interpretation for DNN-based prediction.

Network embedding has been increasingly used in many network analytics applications to generate low-dimensional vector representations, so that many off-the-shelf models can be applied to solve a wide variety of data mining tasks. However, similar to many other machine learning methods, network embedding results remain hard to be understood by users. Each dimension in the embedding space usually does not have any specific meaning, thus it is difficult to comprehend how the embedding instances are distributed in the reconstructed space. In addition, heterogeneous content information may be incorporated into network embedding, so it is challenging to specify which source of information is effective in generating the embedding results. In this paper, we investigate the interpretation of network embedding, aiming to understand how instances are distributed in embedding space, as well as explore the factors that lead to the embedding results. We resort to the post-hoc interpretation scheme, so that our approach can be applied to different types of embedding methods. Specifically, the interpretation of network embedding is presented in the form of a taxonomy. Effective objectives and corresponding algorithms are developed towards building the taxonomy. We also design several metrics to evaluate interpretation results. Experiments on real-world datasets from different domains demonstrate that, by comparing with the state-of-the-art alternatives, our approach produces effective and meaningful interpretation to embedding results.

Machine learning (ML) systems have been increasingly applied in web security applications such as spammer detection, malware detection and fraud detection. These applications have an intrinsic adversarial nature where intelligent attackers can adaptively change their behaviors to avoid being detected by the deployed detectors. Existing efforts against adversaries are usually limited by the type of applied ML models or the specific applications such as image classification. Additionally, the working mechanisms of ML models usually cannot be well understood by users, which in turn impede them from understanding the vulnerabilities of models nor improving their robustness. To bridge the gap, in this paper, we propose to investigate whether model interpretation could potentially help adversarial detection. Specifically, we develop a novel adversary-resistant detection framework by utilizing the interpretation of ML models. The interpretation process explains the mechanism of how the target ML model makes prediction for a given instance, thus providing more insights for crafting adversarial samples. The robustness of detectors is then improved through adversarial training with the adversarial samples. A data-driven method is also developed to empirically estimate costs of adversaries in feature manipulation. Our approach is model-agnostic and can be applied to various types of classification models. Our experimental results on two real-world datasets demonstrate the effectiveness of interpretation-based attacks and how estimated feature manipulation cost would affect the behavior of adversaries.

Attributed network embedding has been widely used in modeling real-world systems. The obtained low-dimensional vector representations of nodes preserve their proximity in terms of both network topology and node attributes, upon which different analysis algorithms can be applied. Recent advances in explanation-based learning and human-in-the-loop models show that by involving experts, the performance of many learning tasks can be enhanced. It is because experts have a better cognition in the latent information such as domain knowledge, conventions, and hidden relations. It motivates us to employ experts to transform their meaningful cognition into concrete data to advance network embedding. However, learning and incorporating the expert cognition into the embedding remains a challenging task. Because expert cognition does not have a concrete form, and is difficult to be measured and laborious to obtain. Also, in a real-world network, there are various types of expert cognition such as the comprehension of word meaning and the discernment of similar nodes. It is nontrivial to identify the types that could lead to a significant improvement in the embedding. In this paper, we study a novel problem of exploring expert cognition for attributed network embedding and propose a principled framework NEEC. We formulate the process of learning expert cognition as a task of asking experts a number of concise and general queries. Guided by the exemplar theory and prototype theory in cognitive science, the queries are systematically selected and can be generalized to various real-world networks. The returned answers from the experts contain their valuable cognition. We model them as new edges and directly add into the attributed network, upon which different embedding methods can be applied towards a more informative embedding representation. Experiments on real-world datasets verify the effectiveness and efficiency of NEEC.

Social media provides a platform for seeking information from a large user base. Information seeking in social media, however, occurs simultaneously with users expressing their viewpoints by making statements. Rhetorical questions have the form of a question but serve the function of a statement and are an important tool employed by users to express their viewpoints. Therefore, rhetorical questions might mislead platforms assisting information seeking in social media. It becomes difficult to identify rhetorical questions as they are not syntactically different from other questions. In this article, we develop a framework to identify rhetorical questions by modeling some motivations of the users to post them. We focus on two motivations of the users drawing from linguistic theories to implicitly convey a message and to modify the strength of a statement previously made. We develop a quantitative framework from these motivations to identify rhetorical questions in social media. We evaluate the framework using two datasets of questions posted on a social media platform Twitter and demonstrate its effectiveness in identifying rhetorical questions. This is the first framework, to the best of our knowledge, to model the possible motivations for posting rhetorical questions to identify them on social media platforms.

Network analysis has been widely applied in many real-world tasks, such as gene analysis and targeted marketing. To extract effective features for these analysis tasks, network embedding automatically learns a low-dimensional vector representation for each node, such that the meaningful topological proximity is well preserved. While the embedding algorithms on pure topological structure have attracted considerable attention, in practice, nodes are often abundantly accompanied with other types of meaningful information, such as node attributes, second-order proximity, and link directionality. A general framework for incorporating the heterogeneous information into network embedding could be potentially helpful in learning better vector representations. However, it remains a challenging task to jointly embed the geometrical structure and a distinct type of information due to the heterogeneity. In addition, the real-world networks often contain a large number of nodes, which put demands on the scalability of the embedding algorithms. To bridge the gap, in this article, we propose a general embedding framework named Heterogeneous Information Learning in Large-scale networks (HILL) to accelerate the joint learning. It enables the simultaneous node proximity assessing process to be done in a distributed manner by decomposing the complex modeling and optimization into many simple and independent sub-problems. We validate the significant correlation between the heterogeneous information and topological structure, and illustrate the generalizability of HILL by applying it to perform attributed network embedding and second-order proximity learning. A variation is proposed for link directionality modeling. Experimental results on real-world networks demonstrate the effectiveness and efficiency of HILL.

Intelligent systems built upon complex machine learning and data mining models (e.g., deep neural networks) have shown superior performances on various real-world applications. However, their effectiveness is limited by the difficulty in interpreting the resultant prediction mechanisms or how the results are obtained. In contrast, the results of many simple or shallow models, such as rule-based or tree-based methods, are explainable but not sufficiently accurate. Model interpretability enables the systems to be clearly understood, properly trusted, effectively managed and widely adopted by end users. Interpretations are necessary in applications such as medical diagnosis, fraud detection and object recognition where valid reasons would be significantly helpful, if not necessary, before taking actions based on predictions. This workshop is about interpreting the prediction mechanisms or results of the complex computational models for data mining by taking advantage of simple models which are easier to understand. We wish to exchange ideas on recent approaches to the challenges of model interpretability, identify emerging fields of applications for such techniques, and provide opportunities for relevant interdisciplinary research or projects.

Network embedding leverages the node proximity manifested to learn a low-dimensional node vector representation for each node in the network. The learned embeddings could advance various learning tasks such as node classification, network clustering, and link prediction. Most, if not all, of the existing works, are overwhelmingly performed in the context of plain and static networks. Nonetheless, in reality, network structure often evolves over time with addition/deletion of links and nodes. Also, a vast majority of real-world networks are associated with a rich set of node attributes, and their attribute values are also naturally changing, with the emerging of new content patterns and the fading of old content patterns. These changing characteristics motivate us to seek an effective embedding representation to capture network and attribute evolving patterns, which is of fundamental importance for learning in a dynamic environment. To our best knowledge, we are the first to tackle this problem with the following two challenges: (1) the inherently correlated network and node attributes could be noisy and incomplete, it necessitates a robust consensus representation to capture their individual properties and correlations; (2) the embedding learning needs to be performed in an online fashion to adapt to the changes accordingly. In this paper, we tackle this problem by proposing a novel dynamic attributed network embedding framework - DANE. In particular, DANE first provides an offline method for a consensus embedding and then leverages matrix perturbation theory to maintain the freshness of the end embedding results in an online manner. We perform extensive experiments on both synthetic and real attributed networks to corroborate the effectiveness and efficiency of the proposed framework.

Attributed networks are pervasive in different domains, ranging from social networks, gene regulatory networks to financial transaction networks. This kind of rich network representation presents challenges for anomaly detection due to the heterogeneity of two data representations. A vast majority of existing algorithms assume certain properties of anomalies are given a prior. Since various types of anomalies in real-world attributed networks co-exist, the assumption that priori knowledge regarding anomalies is available does not hold. In this paper, we investigate the problem of anomaly detection in attributed networks generally from a residual analysis perspective, which has been shown to be effective in traditional anomaly detection problems. However, it is a non-trivial task in attributed networks as interactions among instances complicate the residual modeling process. Methodologically, we propose a learning framework to characterize the residuals of attribute information and its coherence with network information for anomaly detection. By learning and analyzing the residuals, we detect anomalies whose behaviors are singularly different from the majority. Experiments on real datasets show the effectiveness and generality of the proposed framework.

Attributed networks, in which network connectivity and node attributes are available, have been increasingly used to model real-world information systems, such as social media and e-commerce platforms. While outlier detection has been extensively studied to identify anomalies that deviate from certain chosen background, existing algorithms cannot be directly applied on attributed networks due to the heterogeneous types of information and the scale of real-world data. Meanwhile, it has been observed that local anomalies, which may align with global condition, are hard to be detected by existing algorithms with interpretability. Motivated by the observations, in this paper, we propose to study the problem of effective and efficient local anomaly detection in attributed networks. In particular, we design a collective way for modeling heterogeneous network and attribute information, and develop a novel and efficient distributed optimization algorithm to handle large-scale data. In the experiments, we compare the proposed framework with the state-of-the-art methods on both real and synthetic datasets, and demonstrate its effectiveness and efficiency through quantitative evaluation and case studies.

Tensor completion has become an effective computational tool in many real-world data-driven applications. Beyond traditional static setting, with the increasing popularity of high velocity streaming data, it requires efficient online processing without reconstructing the whole model from scratch. Existing work on streaming tensor completion is usually built upon the assumption that tensors only grow in one mode. Unfortunately, the assumption does not hold in many real-world situations in which tensors may grow in multiple modes, i.e., multi-aspect streaming tensors. Efficiently modeling and completing these incremental tensors without sacrificing its effectiveness remains a challenging task due to the uncertainty of tensor mode changes and complex data structure of multi-aspect streaming tensors. To bridge this gap, we propose a Multi-Aspect Streaming Tensor completion framework (MAST) based on CANDECOMP/PARAFAC (CP) decomposition to track the subspace of general incremental tensors for completion. In addition, we investigate a special situation where time is one mode of the tensors, and leverage its extra structure information to improve the general framework towards higher effectiveness. Experimental results on four datasets collected from various real-world applications demonstrate the effectiveness and efficiency of the proposed framework.

Network embedding is to learn low-dimensional vector representations for nodes in a network. It has shown to be effective in a variety of tasks such as node classification and link prediction. While embedding algorithms on pure networks have been intensively studied, in many real-world applications, nodes are often accompanied with a rich set of attributes or features, aka attributed networks. It has been observed that network topological structure and node attributes are often strongly correlated with each other. Thus modeling and incorporating node attribute proximity into network embedding could be potentially helpful, though non-trivial, in learning better vector representations. Meanwhile, real-world networks often contain a large number of nodes and features, which put demands on the scalability of embedding algorithms. To bridge the gap, in this paper, we propose an accelerated attributed network embedding algorithm AANE, which enables the joint learning process to be done in a distributed manner by decomposing the complex modeling and optimization into many sub-problems. Experimental results on several real-world datasets demonstrate the effectiveness and efficiency of the proposed algorithm.

The explosive use of social media, in information dissemination and communication, has also made it a popular platform for the spread of rumors. Rumors could be easily propagated and received by a large number of users in social media, resulting in catastrophic effects in the physical world in a very short period. It is a challenging task, if not impossible, to apply classical supervised learning methods to the early detection of rumors, since the labeling process is time-consuming and labor-intensive. Motivated by the fact that abundant label information of historical rumors is publicly available, in this paper, we propose to investigate whether knowledge learned from historical data could potentially help identify newly emerging rumors. In particular, since a disputed factual claim arouses certain reactions such as curiosity, skepticism, and astonishment, we identify and utilize patterns from prior labeled data to help reveal emergent rumors. Experimental results on real-world data sets demonstrate the effectiveness. Further experiments are conducted to show how much earlier it can detect an emerging rumor than traditional approaches.

User interests and expertise are valuable but often hidden resources on social media. For example, Twitter Lists and LinkedIn's Skill Tags provide a partial perspective on what users are known for (by aggregating crowd tagging knowledge), but the vast majority of users are untagged; their interests and expertise are essentially hidden from important applications such as personalized recommendation, community detection, and expert mining. A natural approach to overcome these limitations is to intelligently learn user topical profiles by exploiting information from multiple, heterogeneous footprints: for instance, Twitter users who post similar hashtags may have similar interests, and YouTube users who upvote the same videos may have similar preferences. And yet identifying "similar" users by exploiting similarity in such a footprint space often provides conflicting evidence, leading to poor-quality user profiles. In this paper, we propose a unified model for learning user topical profiles that simultaneously considers multiple footprints. We show how these footprints can be embedded in a generalized optimization framework that takes into account pairwise relations among all footprints for robustly learning user profiles. Through extensive experiments, we find the proposed model is capable of learning high-quality user topical profiles, and leads to a 10-15% improvement in precision and mean average error versus a cross-triadic factorization state-of-the-art baseline.

Attributed network embedding aims to seek low-dimensional vector representations for nodes in a network, such that original network topological structure and node attribute proximity can be preserved in the vectors. These learned representations have been demonstrated to be helpful in many learning tasks such as network clustering and link prediction. While existing algorithms follow an unsupervised manner, nodes in many real-world attributed networks are often associated with abundant label information, which is potentially valuable in seeking more effective joint vector representations. In this paper, we investigate how labels can be modeled and incorporated to improve attributed network embedding. This is a challenging task since label information could be noisy and incomplete. In addition, labels are completely distinct with the geometrical structure and node attributes. The bewildering combination of heterogeneous information makes the joint vector representation learning more difficult. To address these issues, we propose a novel Label informed Attributed Network Embedding (LANE) framework. It can smoothly incorporate label information into the attributed network embedding while preserving their correlations. Experiments on real-world datasets demonstrate that the proposed framework achieves significantly better performance compared with the state-of-the-art embedding algorithms.

In recent years, deep neural networks have yielded immense success on speech recognition, computer vision and natural language processing. However, the exploration of deep neural networks on recommender systems has received relatively less scrutiny. In this work, we strive to develop techniques based on neural networks to tackle the key problem in recommendation --- collaborative filtering --- on the basis of implicit feedback.

The increasing popularity of social media has encouraged health consumers to share, explore, and validate health and wellness information on social networks, which provide a rich repository of Patient Generated Wellness Data (PGWD). While data-driven healthcare has attracted a lot of attention from academia and industry for improving care delivery through personalized healthcare, limited research has been done on harvesting and utilizing PGWD available on social networks. Recently, representation learning has been widely used in many applications to learn low-dimensional embedding of users. However, existing approaches for representation learning are not directly applicable to PGWD due to its domain nature as characterized by longitudinality, incompleteness, and sparsity of observed data as well as heterogeneity of the patient population. To tackle these problems, we propose an approach which directly learns the embedding from longitudinal data of users, instead of vector-based representation. In particular, we simultaneously learn a low-dimensional latent space as well as the temporal evolution of users in the wellness space. The proposed method takes into account two types of wellness prior knowledge: (1) temporal progression of wellness attributes; and (2) heterogeneity of wellness attributes in the patient population. Our approach scales well to large datasets using parallel stochastic gradient descent. We conduct extensive experiments to evaluate our framework at tackling three major tasks in wellness domain: attribute prediction, success prediction, and community detection. Experimental results on two real-world datasets demonstrate the ability of our approach in learning effective user representations.

Social media plays a major role in helping people affected by natural calamities. These people use social media to request information and help in situations where time is a critical commodity. However, generic social media platforms like Twitter and Facebook are not conducive for obtaining answers promptly. Algorithms to ensure prompt responders for questions in social media have to understand and model the factors affecting their response time. In this paper, we draw from sociological studies on information seeking and organizational behavior to identify users who can provide timely and relevant responses to questions posted on social media. We first draw from these theories to model the future availability and past response behavior of the candidate responders and integrate these criteria with user relevance. We propose a learning algorithm from these criteria to derive optimal rankings of responders for a given question. We present questions posted on Twitter as a form of information seeking activity in social media and use them to evaluate our framework. Our experiments demonstrate that the proposed framework is useful in identifying timely and relevant responders for questions in social media.

Over the last decade, the renaissance of Web technologies has transformed the online world into an application (App) driven society. While the abundant Apps have provided great convenience, their sheer number also leads to severe information overload, making it difficult for users to identify desired Apps. To alleviate the information overloading issue, recommender systems have been proposed and deployed for the App domain. However, existing work on App recommendation has largely focused on one single platform (e.g., smartphones), while it ignores the rich data of other relevant platforms (e.g., tablets and computers). In this article, we tackle the problem of cross-platform App recommendation, aiming at leveraging users’ and Apps’ data on multiple platforms to enhance the recommendation accuracy. The key advantage of our proposal is that by leveraging multiplatform data, the perpetual issues in personalized recommender systems—data sparsity and cold-start—can be largely alleviated. To this end, we propose a hybrid solution, STAR (short for “croSs-plaTform App Recommendation”) that integrates both numerical ratings and textual content from multiple platforms. In STAR, we innovatively represent an App as an aggregation of common features across platforms (e.g., App’s functionalities) and specific features that are dependent on the resided platform. In light of this, STAR can discriminate a user’s preference on an App by separating the user’s interest into two parts (either in the App’s inherent factors or platform-aware features). To evaluate our proposal, we construct two real-world datasets that are crawled from the App stores of iPhone, iPad, and iMac. Through extensive experiments, we show that our STAR method consistently outperforms highly competitive recommendation methods, justifying the rationality of our cross-platform App recommendation proposal and the effectiveness of our solution.

Social media platforms have become the most popular means for users to share what is happening around them. The abundance and growing usage of social media has resulted in a large repository of users' social posts, which provides a stethoscope for inferring individuals' lifestyle and wellness. As users' social accounts implicitly reflect their habits, preferences, and feelings, it is feasible for us to monitor and understand the wellness of users by harvesting social media data towards a healthier lifestyle. As a first step towards accomplishing this goal, we propose to automatically extract wellness events from users' published social contents. Existing approaches for event extraction are not applicable to personal wellness events due to its domain nature characterized by plenty of noise and variety in data, insufficient samples, and inter-relation among events.To tackle these problems, we propose an optimization learning framework that utilizes the content information of microblogging messages as well as the relations between event categories. By imposing a sparse constraint on the learning model, we also tackle the problems arising from noise and variation in microblogging texts. Experimental results on a real-world dataset from Twitter have demonstrated the superior performance of our framework.

Social media has emerged to be a popular platform for people to express their viewpoints on political protests like the Arab Spring. Millions of people use social media to communicate and mobilize their viewpoints on protests. Hence, it is a valuable tool for organizing social movements. However, the mechanisms by which protest affects the population is not known, making it difficult to estimate the number of protestors. In this paper, we are inspired by sociological theories of protest participation and propose a framework to predict from the user's past status messages and interactions whether the next post of the user will be a declaration of protest. Drawing concepts from these theories, we model the interplay between the user's status messages and messages interacting with him over time and predict whether the next post of the user will be a declaration of protest. We evaluate the framework using data from the social media platform Twitter on protests during the recent Nigerian elections and demonstrate that it can effectively predict whether the next post of a user is a declaration of protest.

The pervasive presence of social media greatly enriches online users' social activities, resulting in abundant social relations. Social relations provide an independent source for recommendation, bringing about new opportunities for recommender systems. Exploiting social relations to improve recommendation performance attracts a great amount of attention in recent years. Most existing social recommender systems treat social relations homogeneously and make use of direct connections (or strong dependency connections). However, connections in online social networks are intrinsically heterogeneous and are a composite of various relations. While connected users in online social networks form groups, and users in a group share similar interests, weak dependency connections are established among these users when they are not directly connected. In this paper, we investigate how to exploit the heterogeneity of social relations and weak dependency connections for recommendation. In particular, we employ social dimensions to simultaneously capture heterogeneity of social relations and weak dependency connections, and provide principled ways to model social dimensions, and propose a recommendation framework SoDimRec which incorporates heterogeneity of social relations and weak dependency connections based on social dimensions. Experimental results on real-world data sets demonstrate the effectiveness of the proposed framework. We conduct further experiments to understand the important role of social dimensions in the proposed framework.

While social data is being widely used in various applications such as sentiment analysis and trend prediction, its sheer size also presents great challenges for storing, sharing and processing such data. These challenges can be addressed by data summarization which transforms the original dataset into a smaller, yet still useful, subset. Existing methods find such subsets with objective functions based on data properties such as representativeness or informativeness but do not exploit social contexts, which are distinct characteristics of social data. Further, till date very little work has focused on topic preserving data summarization, despite the abundant work on topic modeling. This is a challenging task for two reasons. First, since topic model is based on latent variables, existing methods are not well-suited to capture latent topics. Second, it is difficult to find such social contexts that provide valuable information for building effective topic-preserving summarization model. To tackle these challenges, in this paper, we focus on exploiting social contexts to summarize social data while preserving topics in the original dataset. We take Twitter data as a case study. Through analyzing Twitter data, we discover two social contexts which are important for topic generation and dissemination, namely (i) CrowdExp topic score that captures the influence of both the crowd and the expert users in Twitter and (ii) Retweet topic score that captures the influence of Twitter users' actions. We conduct extensive experiments on two real-world Twitter datasets using two applications. The experimental results show that, by leveraging social contexts, our proposed solution can enhance topic-preserving data summarization and improve application performance by up to 18%.

Recent years have witnessed the prevalence of networked data in various domains. Among them, a large number of networks are not only topologically structured but also have a rich set of features on nodes. These node features are usually of high dimensionality with noisy, irrelevant and redundant information, which may impede the performance of other learning tasks. Feature selection is useful to alleviate these critical issues. Nonetheless, a vast majority of existing feature selection algorithms are predominantly designed in a static setting. In reality, real-world networks are naturally dynamic, characterized by both topology and content changes. It is desirable to capture these changes to find relevant features tightly hinged with network structure continuously, which is of fundamental importance for many applications such as disaster relief and viral marketing. In this paper, we study a novel problem of time-evolving feature selection for dynamic networks in an unsupervised scenario. Specifically, we propose a TeFS framework by leveraging the temporal evolution property of dynamic networks to update the feature selection results incrementally. Experimental results show the superiority of TeFS over the state-of-the-art batch-mode unsupervised feature selection algorithms.

Feature selection has shown its effectiveness to prepare high-dimensional data for many data mining and machine learning tasks. Traditional feature selection algorithms are mainly based on the assumption that data instances are independent and identically distributed. However, this assumption is invalid in networked data since instances are not only associated with high dimensional features but also inherently interconnected with each other. In addition, obtaining label information for networked data is time consuming and labor intensive. Without label information to direct feature selection, it is difficult to assess the feature relevance. In contrast to the scarce label information, link information in networks are abundant and could help select relevant features. However, most networked data has a lot of noisy links, resulting in the feature selection algorithms to be less effective. To address the above mentioned issues, we propose a robust unsupervised feature selection framework NetFS for networked data, which embeds the latent representation learning into feature selection. Therefore, content information is able to help mitigate the negative effects from noisy links in learning latent representations, while good latent representations in turn can contribute to extract more meaningful features. In other words, both phases could cooperate and boost each other. Experimental results on real-world datasets demonstrate the effectiveness of the proposed framework.

Social media is increasingly being used to access and disseminate information on sociopolitical issues like gun rights and general elections. The popularity and openness of social media makes it conducive for some individuals, known as advocates, who use social media to push their agendas on these issues strategically. Identifying these advocates will caution social media users before reading their information and also enable campaign managers to identify advocates for their digital political campaigns. A significant challenge in identifying advocates is that they employ nuanced strategies to shape user opinion and increase the spread of their messages, making it difficult to distinguish them from random users posting on the campaign. In this paper, we draw from social movement theories and design a quantitative framework to study the nuanced message strategies, propagation strategies, and community structure adopted by advocates for political campaigns in social media. Based on observations of their social media activities manifesting from these strategies, we investigate how to model these strategies for identifying them. We evaluate the framework using two datasets from Twitter, and our experiments demonstrate its effectiveness in identifying advocates for political campaigns with ramifications of this work directed towards assisting users as they navigate through social media spaces.

Relational learning has been proposed to cope with the interdependency among linked instances in social network analysis, which often adopts network connectivity and social media content for prediction. A common assumption in existing relational learning methods is that data instances are equally important. The algorithms developed based on the assumption may be significantly affected by outlier data and thus less robust. In the meantime, it has been well established in social sciences that actors are naturally of different social status in a social network. Motivated by findings from social sciences, in this paper, we investigate whether social status analysis could facilitate relational learning. Particularly, we propose a novel framework RESA to model social status using the network structure. It extracts robust and intrinsic latent social dimensions for social actors, which are further exploited as features for supervised learning models. The proposed method is applicable for real-world relational learning problems where noise exists. Extensive experiments are conducted on datasets obtained from real-world social media platforms. Empirical results demonstrate the effectiveness of RESA and further experiments are conducted to help understand the effects of parameter settings to the proposed model and how local social status works.

Image location prediction is used to estimate the geolocation where an image is taken, which is important for many image applications, such as image retrieval, image browsing, and organization. Since a social image contains heterogeneous contents, such as visual content and textual content, effectively incorporating these contents to predict location is nontrivial. Moreover, it is observed that image content patterns and the locations where they may appear correlate hierarchically. Traditional image location prediction methods mainly adopt a single-level architecture and assume images are independently distributed in geographical space, which is not directly adaptable to the hierarchical correlation. In this paper, we propose a geographically hierarchical bi-modal deep belief network (GH-BDBN) model, which is a compositional learning architecture that integrates multi-modal deep learning model with a non-parametric hierarchical prior model. GH-BDBN learns a joint representation capturing the correlations among different types of image content using a bi-modal DBN, with a geographically hierarchical prior over the joint representation to model the hierarchical correlation between image content and location. Then, an efficient inference algorithm is proposed to learn the parameters and the geographical hierarchical structure of geographical locations. Experimental results demonstrate the superiority of our model for image location prediction.

The rapid urban expansion has greatly extended the physical boundary of users' living area and developed a large number of POIs (points of interest). POI recommendation is a task that facilitates users' urban exploration and helps them filter uninteresting POIs for decision making. While existing work of POI recommendation on location-based social networks (LBSNs) discovers the spatial, temporal, and social patterns of user check-in behavior, the use of content information has not been systematically studied. The various types of content information available on LBSNs could be related to different aspects of a user's check-in action, providing a unique opportunity for POI recommendation. In this work, we study the content information on LBSNs w.r.t. POI properties, user interests, and sentiment indications. We model the three types of information under a unified POI recommendation framework with the consideration of their relationship to check-in actions. The experimental results exhibit the significance of content information in explaining user behavior, and demonstrate its power to improve POI recommendation performance on LBSNs.

The rise of social media provides a great opportunity for people to reach out to their social connections to satisfy their information needs. However, generic social media platforms are not explicitly designed to assist information seeking of users. In this paper, we propose a novel framework to identify the social connections of a user able to satisfy his information needs. The information need of a social media user is subjective and personal, and we investigate the utility of his social context to identify people able to satisfy it. We present questions users post on Twitter as instances of information seeking activities in social media. We infer soft community memberships of the asker and his social connections by integrating network and content information. Drawing concepts from the social foci theory, we identify answerers whose community memberships in the question domain overlap with that of the asker. Our experiments demonstrate that the framework is effective in identifying answerers to social media questions.

Studying the bursty nature of cascades in social media is practically important in many applications such as product sales prediction, disaster relief, and stock market prediction. Although the cascade volume prediction has been extensively studied, how to predict when a burst will come remains an open problem. It is challenging to predict the time of the burst due to the ``quick rise and fall'' pattern and the diverse time spans of the cascades. To this end, this paper proposes a classification based approach for burst time prediction by utilizing and modeling rich knowledge in information diffusion. Particularly, we first propose a time window based approach to predict in which time window the burst will appear. This paves the way to transform the time prediction task to a classification problem. To address the challenge that the original time series data of the cascade popularity only are not sufficient for predicting cascades with diverse magnitudes and time spans, we explore rich information diffusion related knowledge and model them in a scale-independent manner. Extensive experiments on a Sina Weibo reposting dataset demonstrate the superior performance of the proposed approach in accurately predicting the burst time of posts.

Social media has become a part of our daily life and we use it for many reasons. One of its uses is to get our questions answered. Given a multitude of social media sites, however, one immediate challenge is to pick the most relevant site for a question. This is a challenging problem because (1) questions are usually short, and (2) social media sites evolve. In this work, we propose to utilize topic specialization to find the most relevant social media site for a given question. In particular, semantic knowledge is considered for topic specialization as it can not only make a question more specific, but also dynamically represent the content of social sites, which relates a given question to a social media site. Thus, we propose to rank social media sites based on combined search engine query results. Our algorithm yields compelling results for providing a meaningful and consistent site recommendation. This work helps further understand the innate characteristics of major social media platforms for the design of social Q&A systems.

The explosive growth of social media sites brings about massive amounts of high-dimensional data. Feature selection is effective in preparing high-dimensional data for data analytics. The characteristics of social media present novel challenges for feature selection. First, social media data is not fully structured and its features are usually not predefined, but are generated dynamically. For example, in Twitter, slang words (features) are created everyday and quickly become popular within a short period of time. It is hard to directly apply traditional batch-mode feature selection methods to find such features. Second, given the nature of social media, label information is costly to collect. It exacerbates the problem of feature selection without knowing feature relevance. On the other hand, opportunities are also unequivocally present with additional data sources; for example, link information is ubiquitous in social media and could be helpful in selecting relevant features. In this paper, we study a novel problem to conduct unsupervised streaming feature selection for social media data. We investigate how to exploit link information in streaming feature selection, resulting in a novel unsupervised streaming feature selection framework USFS. Experimental results on two real-world social media datasets show the effectiveness and efficiency of the proposed framework comparing with the state-of-the-art unsupervised feature selection algorithms.

Social media is being increasingly used to request information and help in situations like natural disasters, where time is a critical commodity. However, generic social media platforms are not explicitly designed for timely information seeking, making it difficult for users to obtain prompt responses. Algorithms to ensure prompt responders for questions in social media have to understand the factors affecting their response time. In this paper, we draw from sociological studies on information seeking and organizational behavior to model the future availability and past response behavior of the candidate responders. We integrate these criteria with their interests to identify users who can provide timely and relevant responses to questions posted in social media. We propose a learning algorithm to derive optimal rankings of responders for a given question. We present questions posted on Twitter as a form of information seeking activity in social media. Our experiments demonstrate that the proposed framework is useful in identifying timely and relevant responders for questions in social media.

The explosive popularity of microblogging services produce a large volume of microblogging messages. It presents great difficulties for a user to quickly gauge his/her followees' opinions when the user interface is overwhelmed by a large number of messages. Useful information is buried in disorganized, incomplete, and unstructured text messages. We propose to organize the large amount of messages into clusters with meaningful cluster labels, thus provide an overview of the content to fulfill users' information needs. Clustering and labeling of microblogging messages are challenging because that the length of the messages are much shorter than conventional text documents. They usually cannot provide sufficient term co-occurrence information for capturing their semantic associations. As a result, traditional text representation models tend to yield unsatisfactory performance. In this paper, we present a text representation framework by harnessing the power of semantic knowledge bases, i.e., Wikipedia and Wordnet. The originally uncorrelated texts are connected with the semantic representation, thus it enhances the performance of short text clustering and labeling. The experimental results on Twitter and Facebook datasets demonstrate the superior performance of our framework in handling noisy and short microblogging messages.

The explosive use of social media also makes it a popular platform for malicious users, known as social spammers, to overwhelm normal users with unwanted content. One effective way for social spammer detection is to build a classifier based on content and social network information. However, social spammers are sophisticated and adaptable to game the system with fast evolving content and network patterns. First, social spammers continually change their spamming content patterns to avoid being detected. Second, reflexive reciprocity makes it easier for social spammers to establish social influence and pretend to be normal users by quickly accumulating a large number of "human" friends. It is challenging for existing anti-spamming systems based on batch-mode learning to quickly respond to newly emerging patterns for effective social spammer detection. In this paper, we present a general optimization framework to collectively use content and network information for social spammer detection, and provide the solution for efficient online processing. Experimental results on Twitter datasets confirm the effectiveness and efficiency of the proposed framework.

Trust plays a crucial role in helping users collect reliable information in an online world, and has attracted more and more attention in research communities lately. As a conceptual counterpart of trust, distrust can be as important as trust. However, distrust is rarely studied in social media because distrust information is usually unavailable. The value of distrust has been widely recognized in social sciences and recent work shows that distrust can benefit various online applications in social media. In this work, we investigate whether we can obtain distrust information via learning when it is not directly available, and propose to study a novel problem - predicting distrust using pervasively available interaction data in an online world. In particular, we analyze interaction data, provide a principled way to mathematically incorporate interaction data in a novel framework dTrust to predict distrust information. Experimental results using real-world data show that distrust information is predictable with interaction data by the proposed framework dTrust. Further experiments are conducted to gain a deep understand on which factors contribute to the effectiveness of the proposed framework.

Social media is a popular platform for spammers to unfairly overwhelm normal users with unwanted or fake content via social networking. The spammers significantly hinder the use of social media systems for effective information dissemination and sharing. Different from the spammers in traditional platforms such as email and the Web, spammers in social media can easily connect with each other, sometimes without mutual consent. They collude with each other to imitate normal users by quickly accumulating a large number of "human" friends. In addition, content information in social media is noisy and unstructured. It is infeasible to directly apply traditional spammer detection methods in social media. Understanding and detecting deception has been extensively studied in traditional sociology and social sciences. Motivated by psychological findings in physical world, we investigate whether sentiment analysis can help spammer detection in online social media. In particular, we first conduct an exploratory study to analyze the sentiment differences between spammers and normal users, and then present an optimization formulation that incorporates sentiment information into a novel social spammer detection framework. Experimental results on real-world social media datasets show the superior performance of the proposed framework by harnessing sentiment analysis for social spammer detection.

Inferring diffusion networks from traces of cascades has been extensively studied to better understand information diffusion in many domains. A widely used assumption in previous work is that the diffusion network is homogenous and diffusion processes of cascades follow the same pattern. However, in social media, users may have various interests and the connections among them are usually multi-faceted. In addition, different cascades normally diffuse at different speeds and spread to diverse scales, and hence show various diffusion patterns. It is challenging for traditional models to capture the heterogeneous user interactions and diverse patterns of cascades in social media. In this paper, we investigate a novel problem of inferring multi-aspect diffusion networks with multi-pattern cascades. In particular, we study the effects of various diffusion patterns on the information diffusion process by analyzing users' retweeting behavior on a microblogging dataset. By incorporating aspect-level user interactions and various diffusion patterns, a new model for inferring Multi-aspect transmission Rates between users using Multi-pattern cascades (MMRate) is proposed. We also provide an Expectation Maximization algorithm to effectively estimate the parameters. Experimental results on both synthetic and microblogging datasets demonstrate the superior performance of our approach over the state-of-the-art methods in inferring multi-aspect diffusion networks.

Feature selection has been proven to be efficient in preparing high dimensional data for data mining and machine learning. As most data is unlabeled, unsupervised feature selection has attracted more and more attention in recent years. Discriminant analysis has been proven to be a powerful technique to select discriminative features for supervised feature selection. To apply discriminant analysis, we usually need label information which is absent for unlabeled data. This gap makes it challenging to apply discriminant analysis for unsupervised feature selection. In this paper, we investigate how to exploit discriminant analysis in unsupervised scenarios to select discriminative features. We introduce the concept of pseudo labels, which enable discriminant analysis on unlabeled data, propose a novel unsupervised feature selection framework DisUFS which incorporates learning discriminative features with generating pseudo labels, and develop an effective algorithm for DisUFS. Experimental results on different types of real-world data demonstrate the effectiveness of the proposed framework DisUFS.

While microblogging has emerged as an important information sharing and communication platform, it has also become a convenient venue for spammers to overwhelm other users with unwanted content. Currently, spammer detection in microblogging focuses on using social networking information, but little on content analysis due to the distinct nature of microblogging messages. First, label information is hard to obtain. Second, the texts in microblogging are short and noisy. As we know, spammer detection has been extensively studied for years in various media, e.g., emails, SMS and the web. Motivated by abundant resources available in the other media, we investigate whether we can take advantage of the existing resources for spammer detection in microblogging. While people accept that texts in microblogging are different from those in other media, there is no quantitative analysis to show how different they are. In this paper, we first perform a comprehensive linguistic study to compare spam across different media. Inspired by the findings, we present an optimization formulation that enables the design of spammer detection in microblogging using knowledge from external media. We conduct experiments on real-world Twitter datasets to verify (1) whether email, SMS and web spam resources help and (2) how different media help for spammer detection in microblogging.

Recent years have witnessed the tremendous development of social media, which attracts a vast number of Internet users. The high-dimension content generated by these users provides an unique opportunity to understand their behavior deeply. As one of the most fundamental topics, location estimation attracts more and more research efforts. Different from the previous literature, we find that user's location is strongly related to user interest. Based on this, we first build a detection model to mine user interest from short text. We then establish the mapping between location function and user interest before presenting an efficient framework to predict the user's location with convincing fidelity. Thorough evaluations and comparisons on an authentic data set show that our proposed model significantly outperforms the state-of-the-arts approaches. Moreover, the high efficiency of our model also guarantees its applicability in real-world scenarios.

The rapid growth of location-based social networks (LBSNs) invigorates an increasing number of LBSN users, providing an unprecedented opportunity to study human mobile behavior from spatial, temporal, and social aspects. Among these aspects, temporal effects offer an essential contextual cue for inferring a user's movement. Strong temporal cyclic patterns have been observed in user movement in LBSNs with their correlated spatial and social effects (i.e., temporal correlations). It is a propitious time to model these temporal effects (patterns and correlations) on a user's mobile behavior. In this paper, we present the first comprehensive study of temporal effects on LBSNs. We propose a general framework to exploit and model temporal cyclic patterns and their relationships with spatial and social data. The experimental results on two real-world LBSN datasets validate the power of temporal effects in capturing user mobile behavior, and demonstrate the ability of our framework to select the most effective location prediction algorithm under various combinations of prediction models.

Location-based social networks (LBSNs) have attracted an inordinate number of users and greatly enriched the urban experience in recent years. The availability of spatial, temporal and social information in online LBSNs offers an unprecedented opportunity to study various aspects of human behavior, and enable a variety of location-based services such as location recommendation. Previous work studied spatial and social influences on location recommendation in LBSNs. Due to the strong correlations between a user's check-in time and the corresponding check-in location, recommender systems designed for location recommendation inevitably need to consider temporal effects. In this paper, we introduce a novel location recommendation framework, based on the temporal properties of user movement observed from a real-world LBSN dataset. The experimental results exhibit the significance of temporal patterns in explaining user behavior, and demonstrate their power to improve location recommendation performance.

Online reviews play a vital role in the decision-making process for online users. Helpful reviews are usually buried in a large number of unhelpful reviews, and with the consistently increasing number of reviews, it becomes more and more difficult for online users to find helpful reviews. Therefore most online review websites allow online users to rate the helpfulness of a review and a global helpfulness score is computed for the review based on its available ratings. However, in reality, user-specified helpfulness ratings for reviews are very sparse - a few reviews attract large numbers of helpfulness ratings while most reviews obtain few or even no helpfulness ratings. The available helpfulness ratings are too sparse for online users to assess the helpfulness of reviews. Also the helpfulness of a review is not necessarily equally useful for all users and users with different background may treat the helpfulness of a review very differently. The user idiosyncracy of review helpfulness motivates us to study the problem of review helpfulness rating prediction in this paper. We first identify various types of context information, model them mathematically, and propose a context-aware review helpfulness rating prediction framework CAP. Experimental results demonstrate the effectiveness of the proposed framework and the importance of context awareness in solving the review helpfulness rating prediction problem.

The explosive popularity of social media produces mountains of high-dimensional data and the nature of social media also determines that its data is often unla-belled, noisy and partial, presenting new challenges to feature selection. Social media data can be represented by heterogeneous feature spaces in the form of multiple views. In general, multiple views can be complementary and, when used together, can help handle noisy and partial data for any single-view feature selection. These unique challenges and properties motivate us to develop a novel feature selection framework to handle multi-view social media data. In this paper, we investigate how to exploit relations among views to help each other select relevant features, and propose a novel unsupervised feature selection framework, MVFS, for multiview social media data. We systematically evaluate the proposed framework in multi-view datasets from social media websites and the results demonstrate the effectiveness and potential of MVFS.

Supervised learning, e.g., classification, plays an important role in processing and organizing microblogging data. In microblogging, it is easy to mass vast quantities of unlabeled data, but would be costly to obtain labels, which are essential for supervised learning algorithms. In order to reduce the labeling cost, active learning is an effective way to select representative and informative instances to query for labels for improving the learned model. Different from traditional data in which the instances are assumed to be independent and identically distributed (i.i.d.), instances in microblogging are networked with each other. This presents both opportunities and challenges for applying active learning to microblogging data. Inspired by social correlation theories, we investigate whether social relations can help perform effective active learning on networked data. In this paper, we propose a novel Active learning framework for the classification of Networked Texts in microblogging (ActNeT). In particular, we study how to incorporate network information into text content modeling, and design strategies to select the most representative and informative instances from microblogging for labeling by taking advantage of social network structure. Experimental results on Twitter datasets show the benefit of incorporating network information in active learning and that the proposed framework outperforms existing state-of-the-art methods.

Microblogging, like Twitter and Sina Weibo, has become a popular platform of human expressions, through which users can easily produce content on breaking news, public events, or products. The massive amount of microblogging data is a useful and timely source that carries mass sentiment and opinions on various topics. Existing sentiment analysis approaches often assume that texts are independent and identically distributed (i.i.d.), usually focusing on building a sophisticated feature space to handle noisy and short texts, without taking advantage of the fact that the microblogs are networked data. Inspired by the social sciences findings that sentiment consistency and emotional contagion are observed in social networks, we investigate whether social relations can help sentiment analysis by proposing a Sociological Approach to handling Noisy and short Texts (SANT) for sentiment classification. In particular, we present a mathematical optimization formulation that incorporates the sentiment consistency and emotional contagion theories into the supervised learning process; and utilize sparse learning to tackle noisy texts in microblogging. An empirical study of two real-world Twitter datasets shows the superior performance of our framework in handling noisy and short tweets.

Trust plays a crucial role for online users who seek reliable information. However, in reality, user-specified trust relations are very sparse, i.e., a tiny number of pairs of users with trust relations are buried in a disproportionately large number of pairs without trust relations, making trust prediction a daunting task. As an important social concept, however, trust has received growing attention and interest. Social theories are developed for understanding trust. Homophily is one of the most important theories that explain why trust relations are established. Exploiting the homophily effect for trust prediction provides challenges and opportunities. In this paper, we embark on the challenges to investigate the trust prediction problem with the homophily effect. First, we delineate how it differs from existing approaches to trust prediction in an unsupervised setting. Next, we formulate the new trust prediction problem into an optimization problem integrated with homophily, empirically evaluate our approach on two datasets from real-world product review sites, and compare with representative algorithms to gain a deep understanding of the role of homophily in trust prediction.

The explosion of social media services presents a great opportunity to understand the sentiment of the public via analyzing its large-scale and opinion-rich data. In social media, it is easy to amass vast quantities of unlabeled data, but very costly to obtain sentiment labels, which makes unsupervised sentiment analysis essential for various applications. It is challenging for traditional lexicon-based unsupervised methods due to the fact that expressions in social media are unstructured, informal, and fast-evolving. Emoticons and product ratings are examples of emotional signals that are associated with sentiments expressed in posts or words. Inspired by the wide availability of emotional signals in social media, we propose to study the problem of unsupervised sentiment analysis with emotional signals. In particular, we investigate whether the signals can potentially help sentiment analysis by providing a unified way to model two main categories of emotional signals, i.e., emotion indication and emotion correlation. We further incorporate the signals into an unsupervised learning framework for sentiment analysis. In the experiment, we compare the proposed framework with the state-of-the-art methods on two Twitter datasets and empirically evaluate our proposed framework to gain a deep understanding of the effects of emotional signals.

The volume of microblogging messages is increasing exponentially with the popularity of microblogging services. With a large number of messages appearing in user interfaces, it hinders user accessibility to useful information buried in disorganized, incomplete, and unstructured text messages. In order to enhance user accessibility, we propose to aggregate related microblogging messages into clusters and automatically assign them semantically meaningful labels. However, a distinctive feature of microblogging messages is that they are much shorter than conventional text documents. These messages provide inadequate term co occurrence information for capturing semantic associations. To address this problem, we propose a novel framework for organizing unstructured microblogging messages by transforming them to a semantically structured representation. The proposed framework first captures informative tree fragments by analyzing a parse tree of the message, and then exploits external knowledge bases (Wikipedia and WordNet) to enhance their semantic information. Empirical evaluation on a Twitter dataset shows that our framework significantly outperforms existing state-of-the-art methods.

This paper is concerned with the problem of question recommendation in the setting of Community Question Answering (CQA). Given a question as query, our goal is to rank all of the retrieved questions according to their likelihood of being good recommendations for the query. In this paper, we propose a notion of public interest, and show how public interest can boost the performance of question recommendation. In particular, to model public interest in question recommendation, we build a language model to combine relevance score to the query and popularity score regarding question popularity. Experimental results on Yahoo!Answers dataset demonstrate the performance of question recommendation can be greatly improved with considering the public interest.

Existing question retrieval models work relatively well in finding similar questions in community-based question answering (cQA) services. However, they are designed for single-sentence queries or bag-of-word representations, and are not sufficient to handle multi-sentence questions complemented with various contexts. Segmenting questions into parts that are topically related could assist the retrieval system to not only better understand the user's different information needs but also fetch the most appropriate fragments of questions and answers in cQA archive that are relevant to user's query. In this paper, we propose a graph based approach to segmenting multi-sentence questions. The results from user studies show that our segmentation model outperforms traditional systems in question segmentation by over 30% in user's satisfaction. We incorporate the segmentation model into existing cQA question retrieval framework for more targeted question matching, and the empirical evaluation results demonstrate that the segmentation boosts the question retrieval performance by up to 12.93% in Mean Average Precision and 11.72% in Top One Precision. Our model comes with a comprehensive question detector equipped with both lexical and syntactic features.

Clustering of short texts, such as snippets, presents great challenges in existing aggregated search techniques due to the problem of data sparseness and the complex semantics of natural language. As short texts do not provide sufficient term occurring information, traditional text representation methods, such as ``bag of words" model, have several limitations when directly applied to short texts tasks. In this paper, we propose a novel framework to improve the performance of short texts clustering by exploiting the internal semantics from original text and external concepts from world knowledge. The proposed method employs a hierarchical three-level structure to tackle the data sparsity problem of original short texts and reconstruct the corresponding feature space with the integration of multiple semantic knowledge bases -- Wikipedia and WordNet. Empirical evaluation with Reuters and real web dataset demonstrates that our approach is able to achieve significant improvement as compared to the state-of-the-art methods.