Zehua Pei, Hui-Ling Zhen, Lancheng Zou, Xianzhi Yu, Wulong Liu, Sinno Jialin Pan, Mingxuan Yuan, Bei Yu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

To address distribution shifts between training and test data, domain generalization (DG) leverages multiple source domains to learn a model that generalizes well to unseen domains. However, existing DG methods often overfit to the source domains, partly due to the limited coverage of the expected region in feature space. Motivated by this, we propose performing mixup with data interpolation and extrapolation to cover potentially unseen regions. To prevent the detrimental effects of unconstrained extrapolation, we carefully design a policy to generate the instance weights, named F latness-aware G radient-based Mix up (FGMix). The policy relies on gradient-based compatibilities to assign greater weights to instances that carry more invariant information and learn the mixup policy towards flatter minima for better generalization. On the DomainBed benchmark, we validate the efficacy of various designs of FGMix and demonstrate its superiority over other DG algorithms.

Neural pruning aims to compress and accelerate deep neural networks by identifying the optimal subnetwork within a specified sparsity budget. In this work, we study how to gradually sparsify the unpruned dense model to the target sparsity level with minimal performance drop. Specifically, we analyze the evolution of the population of optimal subnetworks under continuous sparsity increments from a thermodynamic perspective. We first reformulate neural pruning as an expected loss minimization problem over the mask distributions. Then, we establish an effective approximation for the sparsity evolution of the optimal mask distribution, termed the **S**parsity Evolutionary **F**okker-**P**lanck-**K**olmogorov Equation (**SFPK**), which provides closed-form, mathematically tractable guidance on distributional transitions for minimizing the expected loss under an infinitesimal sparsity increment. On top of that, we propose SFPK-pruner, a particle simulation-based probabilistic pruning method, to sample performant masks with desired sparsity from the destination distribution of SFPK. In theory, we establish the convergence guarantee for the proposed SFPK-pruner. Our SFPK-pruner exhibits competitive performance in various pruning scenarios. The code is available on https://github.com/mzf666/SFPK-main.

Pretraining large language models often requires significant computational resources and memory due to their vast parameter amount. An effective approach to enhance parameter efficiency in both training and inference is to parameterize each full-size weight as the product of two trainable low-rank factors. While low-rank fine-tuning has achieved great success, low-rank pretraining remains challenging as it requires learning extensive knowledge from scratch under the restrictive low-rank parameterization. During standard low-rank pretraining, separately optimizing the low-rank factors introduces redundant information from the full gradient, which hinders the learning process. To achieve efficient yet effective low-rank pretraining, we propose a **Lo**w-rank **R**iemannian **O**ptimizer (**LORO**). At each LORO update step, the low-rank factor pairs are jointly updated to ensure their full-size product moves along the steepest descent direction on the low-rank manifold, without the need to compute any memory-intensive full-size matrices or gradients. Hence, our LORO finds low-rank models that achieve high performance comparable to full-size pretrained models, while significantly reducing memory usage and accelerating both training and inference. A LLaMA 1B model pretrained with LORO achieves a perplexity score of 2\% better than the full-size baseline, with a 54\% reduction in model memory, a $\times1.8$ speedup in training, and a $\times2.2$ speedup in inference. The code is available on https://github.com/mzf666/LORO-main.

KV cache quantization can improve Large Language Models (LLMs) inference throughput and latency in long contexts and large batch-size scenarios while preserving LLMs effectiveness. However, current methods have three unsolved issues: overlooking layer-wise sensitivity to KV cache quantization, high overhead of online fine-grained decision-making, and low flexibility to different LLMs and constraints. Therefore, we theoretically analyze the inherent correlation of layer-wise transformer attention patterns to KV cache quantization errors and study why key cache is generally more important than value cache for quantization error reduction. We further propose a simple yet effective framework KVTuner to adaptively search for the optimal hardware-friendly layer-wise KV quantization precision pairs for coarse-grained KV cache with multi-objective optimization and directly utilize the offline searched configurations during online inference. To reduce the computational cost of offline calibration, we utilize the intra-layer KV precision pair pruning and inter-layer clustering to reduce the search space. Experimental results show that we can achieve nearly lossless 3.25-bit mixed precision KV cache quantization for LLMs like Llama-3.1-8B-Instruct and 4.0-bit for sensitive models like Qwen2.5-7B-Instruct on mathematical reasoning tasks. The maximum inference throughput can be improved by 21.25% compared with KIVI-KV8 quantization over various context lengths. Our code and searched configurations are available at https://github.com/cmd2001/KVTuner.

We employ a game-theoretic framework to study the impact of a specific strategic behavior among creators—group behavior—on recommendation platforms. In this setting, creators within a group collaborate to maximize their collective utility. We show that group behavior has a limited effect on the game’s equilibrium when the group size is small. However, when the group size is large, group behavior can significantly alter content distribution and user welfare. Specifically, in a top-$K$ recommendation system with exposure-based rewards, we demonstrate that user welfare can suffer a significant loss due to group strategies, and user welfare does not necessarily increase with larger values of $K$ or more random matching, contrasting sharply with the individual creator case. Furthermore, we investigate user welfare guarantees through the lens of the Price of Anarchy (PoA). In the general case, we establish a negative result on the bound of PoA with exposure rewards, proving that it can be arbitrarily large. We then investigate a user engagement rewarding mechanism, which mitigates the issues caused by large group behavior, showing that $\text{PoA}\leq K+1$ in the general case and $\text{PoA}\leq 2$ in the binary case. Empirical results from simulations further support the effectiveness of the user engagement rewarding mechanism.

The design of biological sequences is essential for engineering functional biomolecules that contribute to advancements in human health and biotechnology. Recent advances in diffusion models, with their generative power and efficient conditional sampling, have made them a promising approach for sequence generation. To enhance model performance on limited data and enable multi-objective design and optimization, reinforcement learning (RL)-based fine-tuning has shown great potential. However, existing post-sampling and fine-tuning methods either lack stability in discrete optimization when avoiding gradients or incur high computational costs when employing gradient-based approaches, creating significant challenges for achieving both control and stability in the tuning process. To address these limitations, we propose GLID$^2$E, a gradient-free RL-based tuning approach for discrete diffusion models. Our method introduces a clipped likelihood constraint to regulate the exploration space and implements reward shaping to better align the generative process with design objectives, ensuring a more stable and efficient tuning process. By integrating these techniques, GLID$^2$E mitigates training instabilities commonly encountered in RL and diffusion-based frameworks, enabling robust optimization even in challenging biological design tasks. In the DNA sequence and protein sequence design systems, GLID$^2$E achieves competitive performance in function-based design while maintaining computational efficiency and a flexible tuning mechanism.

This paper introduces MetaDefense, a novel framework for defending against finetuning-based jailbreak attacks in large language models (LLMs). We observe that existing defense mechanisms fail to generalize to harmful queries disguised by unseen attack templates, despite LLMs being capable of distinguishing disguised harmful queries in the embedding space. Based on these insights, we propose a two-stage defense approach: (i) pre-generation defense that detects harmful queries before response generation begins, and (ii) mid-generation defense that monitors partial responses during generation to prevent outputting more harmful content. Our MetaDefense trains the LLM to predict the harmfulness of both queries and partial responses using specialized prompts, enabling early termination of potentially harmful interactions. Extensive experiments across multiple LLM architectures (LLaMA-2-7B, Qwen-2.5-3B-Instruct, and LLaMA-3.2-3B-Instruct) demonstrate that MetaDefense significantly outperforms existing defense mechanisms, achieving robust defense against harmful queries with seen and unseen attack templates while maintaining competitive performance on benign tasks. Code is available at [https://github.com/ws-jiang/MetaDefense](https://github.com/ws-jiang/MetaDefense).

Unsupervised domain adaptation (UDA) is becoming a prominent solution for the domain-shift problem in many time-series classification tasks. With sequence properties, time-series data contain both local and sequential features, and the domain shift exists in both features. However, conventional UDA methods usually cannot distinguish those two features but mix them into one variable for direct alignment, which harms the performance. To address this problem, we propose a novel virtual-label-based hierarchical domain adaptation (VLH-DA) approach for time-series classification. Specifically, we first slice the original time-series data and introduce virtual labels to represent the type of each slice (called local patterns). With the help of virtual labels, we decompose the end-to-end (i.e., signal to time-series label) time-series task into two parts, i.e., signal sequence to local pattern sequence and local pattern sequence to time-series label. By decomposing the complex time-series UDA task into two simpler subtasks, the local features and sequential features can be aligned separately, making it easier to mitigate distribution discrepancies. Experiments on four public time-series datasets demonstrate that our VLH-DA outperforms all state-of-the-art (SOTA) methods.

Solving the task of inverse imaging problems can restore unknown clean images from input measurements that have incomplete information. Utilizing powerful generative models, such as denoising diffusion models, could better tackle the ill-posed issues of inverse problems with the distribution prior of the unknown clean images. We propose a learnable state-estimator-based diffusion model to incorporate the measurements into the reconstruction process. Our method makes efficient use of the pre-trained diffusion models with computational feasibility compared to the conditional diffusion models, which need to be trained from scratch. In addition, our pipeline does not require explicit knowledge of the image degradation operator or make the assumption of its form, unlike many other works that use the pre-trained diffusion models at the test time. The experiments on three typical inverse imaging problems (both linear and non-linear), inpainting, deblurring, and JPEG compression restoration, have comparable results with the state-of-the-art methods.

While large language models (LLMs) exhibit remarkable capabilities across a wide range of tasks, they pose potential safety concerns, such as the ``jailbreak'' problem, wherein malicious instructions can manipulate LLMs to exhibit undesirable behavior. Although several preventive measures have been developed to mitigate the potential risks associated with LLMs, they have primarily focused on English. In this study, we reveal the presence of multilingual jailbreak challenges within LLMs and consider two potential risky scenarios: unintentional and intentional. The unintentional scenario involves users querying LLMs using non-English prompts and inadvertently bypassing the safety mechanisms, while the intentional scenario concerns malicious users combining malicious instructions with multilingual prompts to deliberately attack LLMs. The experimental results reveal that in the unintentional scenario, the rate of unsafe content increases as the availability of languages decreases. Specifically, low-resource languages exhibit about three times the likelihood of encountering harmful content compared to high-resource languages, with both ChatGPT and GPT-4. In the intentional scenario, multilingual prompts can exacerbate the negative impact of malicious instructions, with astonishingly high rates of unsafe output: 80.92\% for ChatGPT and 40.71\% for GPT-4. To handle such a challenge in the multilingual context, we propose a novel \textsc{Self-Defense} framework that automatically generates multilingual training data for safety fine-tuning. Experimental results show that ChatGPT fine-tuned with such data can achieve a substantial reduction in unsafe content generation. Data is available at \url{https://github.com/DAMO-NLP-SG/multilingual-safety-for-LLMs}.

Graph Neural Networks are popular tools in graph representation learning that capture the graph structural properties. However, most GNNs employ single-resolution graph feature extraction, thereby failing to capture micro-level local patterns (high resolution) and macro-level graph cluster and community patterns (low resolution) simultaneously. Many multiresolution methods have been developed to capture graph patterns at multiple scales, but most of them depend on predefined and handcrafted multiresolution transforms that remain fixed throughout the training process once formulated. Due to variations in graph instances and distributions, fixed handcrafted transforms can not effectively tailor multiresolution representations to each graph instance. To acquire multiresolution representation suited to different graph instances and distributions, we introduce the Multiresolution Meta-Framelet-based Graph Convolutional Network (MM-FGCN), facilitating comprehensive and adaptive multiresolution analysis across diverse graphs. Extensive experiments demonstrate that our MM-FGCN achieves SOTA performance on various graph learning tasks.

Deploying a safe mobile robot policy in scenarios with human pedestrians is challenging due to their unpredictable movements. Current Reinforcement Learningbased motion planners rely on a single policy to simulate pedestrian movements and could suffer from the over-fitting issue. Alternatively, framing the collision avoidance problem as a multi-agent framework, where agents generate dynamic movements while learning to reach their goals, can lead to conflicts with human pedestrians due to their homogeneity.To tackle this problem, we introduce an efficient method that enhances agent diversity within a single policy by maximizing an information-theoretic objective. This diversity enriches each agent’s experiences, improving its adaptability to unseen crowd behaviors. In assessing an agent’s robustness against unseen crowds, we propose diverse scenarios inspired by pedestrian crowd behaviors. Our behavior-conditioned policies outperform existing works in these challenging scenes, reducing potential collisions without additional time or travel.

In reinforcement learning with image-based inputs, it is crucial to establish a robust and generalizable state representation. Recent advancements in metric learning, such as deep bisimulation metric approaches, have shown promising results in learning structured low-dimensional representation space from pixel observations, where the distance between states is measured based on task-relevant features. However, these approaches face challenges in demanding generalization tasks and scenarios with non-informative rewards. This is because they fail to capture sufficient long-term information in the learned representations. To address these challenges, we propose a novel State Chrono Representation (SCR) approach. SCR augments state metric-based representations by incorporating extensive temporal information into the update step of bisimulation metric learning. It learns state distances within a temporal framework that considers both future dynamics and cumulative rewards over current and long-term future states. Our learning strategy effectively incorporates future behavioral information into the representation space without introducing a significant number of additional parameters for modeling dynamics. Extensive experiments conducted in DeepMind Control and Meta-World environments demonstrate that SCR achieves better performance comparing to other recent metric-based methods in demanding generalization tasks. The codes of SCR are available in https://github.com/jianda-chen/SCR.

Generating graph-structured data is a challenging problem, which requires learning the underlying distribution of graphs. Various models such as graph VAE, graph GANs, and graph diffusion models have been proposed to generate meaningful and reliable graphs, among which the diffusion models have achieved state-of-the-art performance. In this paper, we argue that running full-rank diffusion SDEs on the whole graph adjacency matrix space hinders diffusion models from learning graph topology generation, and hence significantly deteriorates the quality of generated graph data. To address this limitation, we propose an efficient yet effective Graph Spectral Diffusion Model (GSDM), which is driven by low-rank diffusion SDEs on the graph spectrum space. Our spectral diffusion model is further proven to enjoy a substantially stronger theoretical guarantee than standard diffusion models. Extensive experiments across various datasets demonstrate that our proposed GSDM turns out to be the SOTA model, by exhibiting both significantly higher generation quality and much less computational consumption than the baselines.

Sequential recommendation systems aim to exploit users’ sequential behavior patterns to capture their interaction intentions and improve recommendation accuracy. Existing sequential recommendation methods mainly focus on modeling the items’ chronological relationships in each individual user behavior sequence, which may not be effective in making accurate and robust recommendations. On the one hand, the performance of existing sequential recommendation methods is usually sensitive to the length of a user’s behavior sequence (i.e., the list of a user’s historically interacted items). On the other hand, besides the context information in each individual user behavior sequence, the collaborative information among different users’ behavior sequences is also crucial to make accurate recommendations. However, this kind of information is usually ignored by existing sequential recommendation methods. In this work, we propose a new sequential recommendation framework, which encodes the context information in each individual user behavior sequence as well as the collaborative information among the behavior sequences of different users, through building a local dependency graph for each item. We conduct extensive experiments to compare the proposed model with state-of-the-art sequential recommendation methods on five benchmark datasets. The experimental results demonstrate that the proposed model is able to achieve better recommendation performance than existing methods, by incorporating collaborative information.

Cross-domain aspect-based sentiment analysis (ABSA) aims to perform various fine-grained sentiment analysis tasks on a target domain by transferring knowledge from a source domain.Since labeled data only exists in the source domain, a model is expected to bridge the domain gap for tackling cross-domain ABSA.Though domain adaptation methods have proven to be effective, most of them are based on a discriminative model, which needs to be specifically designed for different ABSA tasks.To offer a more general solution, we propose a unified bidirectional generative framework to tackle various cross-domain ABSA tasks.Specifically, our framework trains a generative model in both text-to-label and label-to-text directions.The former transforms each task into a unified format to learn domain-agnostic features, and the latter generates natural sentences from noisy labels for data augmentation, with which a more accurate model can be trained.To investigate the effectiveness and generality of our framework, we conduct extensive experiments on four cross-domain ABSA tasks and present new state-of-the-art results on all tasks.Our data and code are publicly available at https://github.com

Sentiment analysis is a well-established natural language processing task, with sentiment polarity classification being one of its most popular and representative tasks.However, despite the success of pre-trained language models in this area, they often fall short of capturing the broader complexities of sentiment analysis.To address this issue, we propose a new task called Sentiment and Opinion Understanding of Language (SOUL).SOUL aims to evaluate sentiment understanding through two subtasks: Review Comprehension (RC) and Justification Generation (JG).RC seeks to validate statements that focus on subjective information based on a review text, while JG requires models to provide explanations for their sentiment predictions.To enable comprehensive evaluation, we annotate a new dataset comprising 15,028 statements from 3,638 reviews.Experimental results indicate that SOUL is a challenging task for both small and large language models, with a performance gap of up to 27% when compared to human performance.Furthermore, evaluations conducted with both human experts and GPT-4 highlight the limitations of the small language model in generating reasoning-based justifications.These findings underscore the challenging nature of the SOUL task for existing models, emphasizing the need for further advancements in sentiment analysis to address its complexities.The new dataset and code are available at https://github.com

Large language models (LLMs) have showcased their capability with few-shot inference known as in-context learning.However, indomain demonstrations are not always readily available in real scenarios, leading to crossdomain in-context learning.Besides, LLMs are still facing challenges in long-tail knowledge in unseen and unfamiliar domains.The above limitations demonstrate the necessity of Unsupervised Domain Adaptation (UDA).In this paper, we study the UDA problem under an in-context learning setting to adapt language models from the source domain to the target domain without any target labels.The core idea is to retrieve a subset of cross-domain elements that are the most similar to the query, and elicit language model to adapt in an in-context manner by learning both target domain distribution and the discriminative task signal simultaneously with the augmented cross-domain incontext examples.We devise different prompting and training strategies, accounting for different LM architectures to learn the target distribution via language modeling.With extensive experiments on Sentiment Analysis (SA) and Named Entity Recognition (NER) tasks, we thoroughly study the effectiveness of ICL for domain transfer and demonstrate significant improvements over baseline models.

We propose a novel progressive parameter-sharing strategy (MPPS) in this paper for effectively training multitask learning models on diverse computer vision tasks simultaneously. Specifically, we propose to parameterize distributions for different tasks to control the sharings, based on the concept of Exclusive Capacity that we introduce. A scheduling mechanism following the concept of curriculum learning is also designed to progressively change the sharing strategy to increase the level of sharing during the learning process. We further propose a novel loss function to regularize the optimization of network parameters as well as the sharing probabilities of each neuron for each task. Our approach can be combined with many state-of-the-art multitask learning solutions to achieve better joint task performance. Comprehensive experiments show that it has competitive performance on three challenging datasets (Multi-CIFAR100, NYUv2, and Cityscapes) using various convolution neural network architectures.

Federated learning (FL) has emerged as a popular machine learning paradigm recently. Compared with traditional distributed learning, its unique challenges mainly lie in communication efficiency and non-IID (heterogeneous data) problem. While the widely adopted framework FedAvg can reduce communication overhead significantly, its effectiveness on non-IID data still lacks exploration. In this paper, we study the non-IID problem of FL from the perspective of domain adaptation. We propose a distribution regularization for FL on non-IID data such that the discrepancy of data distributions between clients is reduced. To further reduce the communication cost, we devise two novel distributed learning algorithms, namely rFedAvg and rFedAvg+, for efficiently learning with the distribution regularization. More importantly, we theoretically establish their convergence for strongly convex objectives. Extensive experiments on 4 datasets with both CNN and LSTM as learning models verify the effectiveness and efficiency of the proposed algorithms.

Neural pruning, which involves identifying the optimal sparse subnetwork, is a key technique for reducing the complexity and improving the efficiency of deep neural networks. To address the challenge of solving neural pruning at a specific sparsity level directly, we investigate the evolution of optimal subnetworks with continuously increasing sparsity, which can provide insight into how to transform an unpruned dense model into an optimal subnetwork with any desired level of sparsity. In this paper, we proposed a novel pruning framework, coined Sparsity-indexed ODE (SpODE) that provides explicit guidance on how to best preserve model performance while ensuring an infinitesimal increase in model sparsity. On top of this, we develop a pruning algorithm, termed Pruning via Sparsity-indexed ODE (PSO), that enables effective pruning via traveling along the SpODE path. Empirical experiments show that PSO achieves either better or comparable performance compared to state-of-the-art baselines across various pruning settings.

Machine learning is becoming increasingly important in today's world. Beyond its powerful performances, there has been an emerging concern about the trustworthiness of machine learning, including but not limited to: robustness to malicious attacks, generalization to unseen datasets, and interpretability to explain its outputs. Such concerns are even more urgent in some safety-critical applications such as medical diagnosis and autonomous driving. Trustworthy machine learning (TrustML) aims to tackle these challenges from the perspectives of theory, algorithm, and applications. In this tutorial, we will give a comprehensive introduction to the recent advance of trustworthy machine learning in robustness, generalization, and interpretability. We will cover their problem formulation, related research, popular algorithms, and successful applications. Additionally, we will also introduce some potential challenges for future research. We do hope that this tutorial will not only serve as a platform to understand TrustML, but also raise the awareness of everyone for more trustworthy applications.

The performance of deep learning models heavily relies on the quality of the training data. Inadequacies in the training data, such as corrupt input or noisy labels, can lead to the failure of model generalization. Recent studies propose repairing the model by identifying the training samples that contribute to the failure and removing their influence from the model. However, it is important to note that the identified data may contain both beneficial and detrimental information. Simply erasing the information of the identified data from the model can have a negative impact on its performance, especially when accurate data is mistakenly identified as detrimental and removed. To overcome this challenge, we propose a novel approach that leverages the knowledge obtained from a retained clean set. Our method first identifies harmful data by utilizing the clean set, then separates the beneficial and detrimental information within the identified data. Finally, we utilize the extracted beneficial information to enhance the model's performance. Through empirical evaluations, we demonstrate that our method outperforms baseline approaches in both identifying harmful data and rectifying model failures. Particularly in scenarios where identification is challenging and a significant amount of benign data is involved, our method improves performance while the baselines deteriorate due to the erroneous removal of beneficial information.

With the availability of massive labeled training data, powerful machine learning models can be trained. However, the traditional I.I.D. assumption that the training and testing data should follow the same distribution is often violated in reality. While existing domain adaptation approaches can tackle domain shift, it relies on the target samples for training. Domain generalization is a promising technology that aims to train models with good generalization ability to unseen distributions. In this tutorial, we will present the recent advance of domain generalization. Specifically, we introduce the background, formulation, and theory behind this topic. Our primary focus is on the methodology, evaluation, and applications. We hope this tutorial can draw interest of the community and provide a thorough review of this area. Eventually, more robust systems can be built for responsible AI. All tutorial materials and updates can be found online at https://dgresearch.github.io/.

To enable effective learning of new tasks with only a few examples, meta-learning acquires common knowledge from the existing tasks with a globally shared meta-learner. To further address the problem of task heterogeneity, recent developments balance between customization and generalization by incorporating task clustering to generate task-aware modulation to be applied to the global meta-learner. However, these methods learn task representation mostly from the features ofinput data, while the task-specific optimization process with respect to the base-learner is often neglected. In this work, we propose a Clustered Task-Aware Meta-Learning (CTML) framework with task representation learned from both features and learning paths. We first conduct rehearsed task learning from the common initialization, and collect a set of geometric quantities that adequately describes this learning path. By inputting this set of values into a meta path learner, we automatically abstract path representation optimized for downstream clustering and modulation. Aggregating the path and feature representations results in an improved task representation. To further improve inference efficiency, we devise a shortcut tunnel to bypass the rehearsed learning process at a meta-testing time. Extensive experiments on two real-world application domains: few-shot image classification and cold-start recommendation demonstrate the superiority of CTML compared to state-of-the-art methods. We provide our code at https://github.com/didiya0825.

Multi-hop reading comprehension requires an ability to reason across multiple documents.On the one hand, deep learning approaches only implicitly encode query-related information into distributed embeddings which fail to uncover the discrete relational reasoning process to infer the correct answer.On the other hand, logic-based approaches provide interpretable rules to infer the target answer, but mostly work on structured data where entities and relations are well-defined.In this paper, we propose a deep-learning based inductive logic reasoning method that firstly extracts query-related (candidate-related) information, and then conducts logic reasoning among the filtered information by inducing feasible rules that entail the target relation.The reasoning process is accomplished via attentive memories with novel differentiable logic operators.To demonstrate the effectiveness of our model, we evaluate it on two reading comprehension datasets, namely WikiHop and MedHop.

With the increasing popularity of GPS modules, there are various urban applications relying on trajectory data modeling. In this work, we study the problem to model the vehicle trajectories by predicting the next road segment given a partial trajectory. Existing methods that model trajectories with Markov chain or recurrent neural network suffer from issues of modeling, context and semantics. In this paper, we propose a new trajectory modeling framework called Multi-task Modeling for Trajectories (MMTraj), which avoids these issues. Specifically, MMTraj uses multi-head self-attention networks for sequential modeling, captures the overall road network as the context information for road segment embedding, and performs an auxiliary task of predicting the trajectory destination to better guide the main trajectory modeling task (controlled by a carefully designed gating mechanism). Extensive experiments conducted on real-world datasets demonstrate the superiority of the proposed method over the baseline methods.

In this work, we study Unsupervised Domain Adaptation (UDA) in a challenging selfsupervised approach.One of the difficulties is how to learn task discrimination in the absence of target labels.Unlike previous literature which directly aligns cross-domain distributions or leverages reverse gradient, we propose Domain Confused Contrastive Learning (DCCL) to bridge the source and the target domains via domain puzzles, and retain discriminative representations after adaptation.Technically, DCCL searches for a most domainchallenging direction and exquisitely crafts domain confused augmentations as positive pairs, then it contrastively encourages the model to pull representations towards the other domain, thus learning more stable and effective domain invariances.We also investigate whether contrastive learning necessarily helps with UDA when performing other data augmentations.Extensive experiments demonstrate that DCCL significantly outperforms baselines.

Learning an informative representation with behavioral metrics is able to accelerate the deep reinforcement learning process. There are two key research issues on behavioral metric-based representation learning: 1) how to relax the computation of a specific behavioral metric, which is difficult or even intractable to compute, and 2) how to approximate the relaxed metric by learning an embedding space for states. In this paper, we analyze the potential relaxation and/or approximation gaps for existing behavioral metric-based representation learning methods. Based on the analysis, we propose a new behavioral distance, the RAP distance, and develop a practical representation learning algorithm on top of it with a theoretical analysis. We conduct extensive experiments on DeepMind Control Suite with distraction, Robosuite, and autonomous driving simulator CARLA to demonstrate new state-of-the-art results.

Linear coupling is recently proposed to accelerate first-order algorithms by linking gradient descent and mirror descent together, which is able to achieve the accelerated convergence rate for first-order algorithms. This work focuses on the convergence analysis of linear coupling for convex composite minimization when the proximal operator cannot be exactly computed. It is of particular interest to study the convergence of linear coupling because it not only achieves the accelerated convergence rate for first-order algorithm but also works for generic norms. We present convergence analysis of linear coupling by allowing the proximal operator to be computed up to a certain precision. Our analysis illustrates that the accelerated convergence rate of linear coupling with inexact proximal operator can be preserved if the error sequence of inexact proximal operator decreases in a sufficiently fast rate. More importantly, our analysis leads to better bounds than existing works on inexact proximal operator. Experiment results on several real-world datasets verify our theoretical results.

The accelerated proximal methods (APM) have become one of the most important optimization tools for large-scale convex composite minimization problems, due to their wide range of applications and the optimal convergence rate in first-order algorithms. However, most existing theoretical results of APM are obtained by assuming that the gradient oracle is exact and the proximal mapping must be exactly solved, which may not hold in practice. This work presents a theoretical study of APM by allowing to use inexact gradient oracle and approximate proximal mapping. Specifically, we analyze inexact APM by improving the approximate duality gap technique (ADGT) which was originally designed for convergence analysis for first-order methods with both exact gradient oracle and proximal mapping. Our approach has several advantages: 1) we provide a unified convergence analysis that allows both inexact gradient oracle and approximate proximal mapping; 2) our proof is generic that naturally recovers the convergence rates of both accelerated and non-accelerated proximal methods, on top of which the advantages and the disadvantages of acceleration can be easily derived; 3) we derive the same convergence bound as previous methods in terms of inexact gradient oracle, but a tighter convergence bound in terms of approximate proximal mapping.

For personalized recommendations, collaborative filtering (CF) methods aim to recommend items to users based on data of historical user-item interactions. Deep learning has indicated success in improving performance of CF methods in recent works. However, to generate an item recommendation list for each user, a lot of deep learning based CF methods require every pair of users and items to be passed through multiple neural layers. This requires intensive computation and makes real-time end-to-end neural recommendations very costly. To address this issue, in this paper, we propose a new deep learning-based hierarchical decision network to filter out irrelevant items to save computation cost while maintaining good recommendation accuracy of deep CF methods. We also develop a distillation-based training algorithm, which uses a well-trained CF model as a teacher network to guide the training of the decision network. We conducted extensive experiments on real-world benchmark datasets to verify the effectiveness of efficiency of our decision network for making recommendations. The experimental results indicate that the proposed decision network is able to maintain or even improve the recommendation quality in terms of various metrics and meanwhile enjoy lower computational cost.

In wearable-sensor-based activity recognition, it is often assumed that the training and test samples follow the same data distribution. This assumption neglects practical scenarios where the activity patterns inevitably vary from person to person. To solve this problem, transfer learning and domain adaptation approaches are often leveraged to reduce the gaps between different participants. Nevertheless, these approaches require additional information (i.e., labeled or unlabeled data, meta-information) from the target domain during the training stage. In this paper, we introduce a novel method named Generalizable Independent Latent Excitation (GILE) for human activity recognition, which greatly enhances the cross-person generalization capability of the model. Our proposed method is superior to existing methods in the sense that it does not require any access to the target domain information. Besides, this novel model can be directly applied to various target domains without re-training or fine-tuning. Specifically, the proposed model learns to automatically disentangle domain-agnostic and domain-specific features, the former of which are expected to be invariant across various persons. To further remove correlations between the two types of features, a novel Independent Excitation mechanism is incorporated in the latent feature space. Comprehensive experimental evaluations are conducted on three benchmark datasets to demonstrate the superiority of the proposed method over the state-of-the-art solutions.

Many challenging partially observable reinforcement learning problems have sparse rewards and most existing model-free algorithms struggle with such reward sparsity. In this paper, we propose a novel reward shaping approach to infer the intrinsic rewards for the agent from a sequential generative model. Specifically, the sequential generative model processes a sequence of partial observations and actions from the agent's historical transitions to compile a belief state for performing forward dynamics prediction. Then we utilize the error of the dynamics prediction task to infer the intrinsic rewards for the agent. Our proposed method is able to derive intrinsic rewards that could better reflect the agent's surprise or curiosity over its ground-truth state by taking a sequential inference procedure. Furthermore, we formulate the inference procedure for dynamics prediction as a multi-step forward prediction task, where the time abstraction that has been incorporated could effectively help to increase the expressiveness of the intrinsic reward signals. To evaluate our method, we conduct extensive experiments on challenging 3D navigation tasks in ViZDoom and DeepMind Lab. Empirical evaluation results show that our proposed exploration method could lead to significantly faster convergence than various state-of-the-art exploration approaches in the testified navigation domains.

Time series has wide applications in the real world and is known to be difficult to forecast. Since its statistical properties change over time, its distribution also changes temporally, which will cause severe distribution shift problem to existing methods. However, it remains unexplored to model the time series in the distribution perspective. In this paper, we term this as Temporal Covariate Shift (TCS). This paper proposes Adaptive RNNs (AdaRNN) to tackle the TCS problem by building an adaptive model that generalizes well on the unseen test data. AdaRNN is sequentially composed of two novel algorithms. First, we propose Temporal Distribution Characterization to better characterize the distribution information in the TS. Second, we propose Temporal Distribution Matching to reduce the distribution mismatch in TS to learn the adaptive TS model. AdaRNN is a general framework with flexible distribution distances integrated. Experiments on human activity recognition, air quality prediction, and financial analysis show that AdaRNN outperforms the latest methods by a classification accuracy of 2.6% and significantly reduces the RMSE by 9.0%. We also show that the temporal distribution matching algorithm can be extended in Transformer structure to boost its performance.

Correlated time series forecasting plays an essential role in many cyber-physical systems, where entities interact with each other over time. To enable accurate forecasting, it is essential to capture both the temporal dynamics and the correlations among different entities. To capture the former, two popular types of models, recurrent neural networks (RNNs) and temporal convolution networks (TCNs), are employed. To capture the latter, a graph is constructed to reflect certain relationships among entities and then graph convolution (GC) is applied upon the graph to capture the correlations among the entities. The state-of-the-art forecasting accuracy is achieved by models that combine RNNs or TCNs with GC. However, they neither capture distinct temporal dynamics that exist among different entities nor consider the entity correlations that evolve across time. In this paper, rather than proposing yet another new end-to-end forecasting model, we aim at providing a framework to enhance existing forecasting models, where we propose generic plugins that can be easily integrated into existing solutions to solve the two challenges and thus further enhance their accuracy. Specifically, we propose two plugin neural networks that are able to better capture distinct temporal dynamics for different entities and dynamic entity correlations across time, so that forecasting accuracy is improved while model parameters to be learned are reduced. Experimental results on three real-world correlated time series data sets demonstrate that the proposed framework with the two plugin networks is able to achieve the above goals.

Attributed event sequences are commonly encountered in practice. A recent research line focuses on incorporating neural networks with the statistical model--marked point processes, which is the conventional tool for dealing with attributed event sequences. Neural marked point processes possess good interpretability of probabilistic models as well as the representational power of neural networks. However, we find that performance of neural marked point processes is not always increasing as the network architecture becomes more complicated and larger, which is what we call the performance saturation phenomenon. This is due to the fact that the generalization error of neural marked point processes is determined by both the network representational ability and the model specification at the same time. Therefore we can draw two major conclusions: first, simple network structures can perform no worse than complicated ones for some cases; second, using a proper probabilistic assumption is as equally, if not more, important as improving the complexity of the network. Based on this observation, we propose a simple graph-based network structure called GCHP, which utilizes only graph convolutional layers, thus it can be easily accelerated by the parallel mechanism. We directly consider the distribution of interarrival times instead of imposing a specific assumption on the conditional intensity function, and propose to use a likelihood ratio loss with a moment matching mechanism for optimization and model selection. Experimental results show that GCHP can significantly reduce training time and the likelihood ratio loss with interarrival time probability assumptions can greatly improve the model performance.

Recommender Systems (RSs) in real-world applications often deal with billions of user interactions daily. To capture the most recent trends effectively, it is common to update the model incrementally using only the newly arrived data. However, this may impede the model’s ability to retain long-term information due to the potential overfitting and forgetting issues. To address this problem, we propose a novel Adaptive Sequential Model Generation (ASMG) framework, which generates a better serving model from a sequence of historical models via a meta generator. For the design of the meta generator, we propose to employ Gated Recurrent Units (GRUs) to leverage its ability to capture the long-term dependencies. We further introduce some novel strategies to apply together with the GRU meta generator, which not only improve its computational efficiency but also enable more accurate sequential modeling. By instantiating the model-agnostic framework on a general deep learning-based RS model, we demonstrate that our method achieves state-of-the-art performance on three public datasets and one industrial dataset.

Abstract Currently, deep learning models have been widely adopted and achieved promising results on various application domains. Despite their intriguing performance, most deep learning models function as black boxes, lacking explicit reasoning capabilities and explanations, which are usually essential for complex problems. Take joint inference in information extraction as an example. This task requires the identification of multiple structured knowledge from texts, which is inter-correlated, including entities, events, and the relationships between them. Various deep neural networks have been proposed to jointly perform entity extraction and relation prediction, which only propagate information implicitly via representation learning. However, they fail to encode the intensive correlations between entity types and relations to enforce their coexistence. On the other hand, some approaches adopt rules to explicitly constrain certain relational facts, although the separation of rules with representation learning usually restrains the approaches with error propagation. Moreover, the predefined rules are inflexible and might result in negative effects when data is noisy. To address these limitations, we propose a variational deep logic network that incorporates both representation learning and relational reasoning via the variational EM algorithm. The model consists of a deep neural network to learn high-level features with implicit interactions via the self-attention mechanism and a relational logic network to explicitly exploit target interactions. These two components are trained interactively to bring the best of both worlds. We conduct extensive experiments ranging from fine-grained sentiment terms extraction, end-to-end relation prediction, to end-to-end event extraction to demonstrate the effectiveness of our proposed method.

Semantic segmentation, which aims to acquire pixel-level understanding about images, is among the key components in computer vision. To train a good segmentation model for real-world images, it usually requires a huge amount of time and labor effort to obtain sufficient pixel-level annotations of real-world images beforehand. To get rid of such a nontrivial burden, one can use simulators to automatically generate synthetic images that inherently contain full pixel-level annotations and use them to train a segmentation model for the real-world images. However, training with synthetic images usually cannot lead to good performance due to the domain difference between the synthetic images (i.e., source domain) and the real-world images (i.e., target domain). To deal with this issue, a number of unsupervised domain adaptation (UDA) approaches have been proposed, where no labeled real-world images are available. Different from those methods, in this work, we conduct a pioneer attempt by using easy-to-collect image-level annotations for target images to improve the performance of cross-domain segmentation. Specifically, we leverage those image-level annotations to construct curriculums for the domain adaptation problem. The curriculums describe multi-level properties of the target domain, including label distributions over full images, local regions and single pixels. Since image annotations are “weak” labels compared to pixel annotations for segmentation, we coin this new problem as weakly-supervised cross-domain segmentation. Comprehensive experiments on the GTA5→ Cityscapes and SYNTHIA→ Cityscapes settings demonstrate the effectiveness of our method over the existing state-of-the-art baselines.

Value iteration networks (VINs) have been demonstrated to have a good generalization ability for reinforcement learning tasks across similar domains. However, based on our experiments, a policy learned by VINs still fail to generalize well on the domain whose action space and feature space are not identical to those in the domain where it is trained. In this paper, we propose a transfer learning approach on top of VINs, termed Transfer VINs (TVINs), such that a learned policy from a source domain can be generalized to a target domain with only limited training data, even if the source domain and the target domain have domain-specific actions and features. We empirically verify that our proposed TVINs outperform VINs when the source and the target domains have similar but not identical action and feature spaces. Furthermore, we show that the performance improvement is consistent across different environments, maze sizes, dataset sizes as well as different values of hyperparameters such as number of iteration and kernel size.

Information extraction (IE) aims to produce structured information from an input text, e.g., Named Entity Recognition and Relation Extraction. Various attempts have been proposed for IE via feature engineering or deep learning. However, most of them fail to associate the complex relationships inherent in the task itself, which has proven to be especially crucial. For example, the relation between 2 entities is highly dependent on their entity types. These dependencies can be regarded as complex constraints that can be efficiently expressed as logical rules. To combine such logic reasoning capabilities with learning capabilities of deep neural networks, we propose to integrate logical knowledge in the form of first-order logic into a deep learning system, which can be trained jointly in an end-to-end manner. The integrated framework is able to enhance neural outputs with knowledge regularization via logic rules, and at the same time update the weights of logic rules to comply with the characteristics of the training data. We demonstrate the effectiveness and generalization of the proposed model on multiple IE tasks.

Though deep learning has achieved significant success in various NLP tasks, most deep learning models lack the capability of encoding explicit domain knowledge to model complex causal relationships among different types of variables.On the other hand, logic rules offer a compact expression to represent the causal relationships to guide the training process.Logic programs can be cast as a satisfiability problem which aims to find truth assignments to logic variables by maximizing the number of satisfiable clauses (MaxSAT).We adopt the MaxSAT semantics to model logic inference process and smoothly incorporate a weighted version of MaxSAT that connects deep neural networks and a graphical model in a joint framework.The joint model feeds deep learning outputs to a weighted MaxSAT layer to rectify the erroneous predictions and can be trained via end-to-end gradient descent.Our proposed model associates the benefits of highlevel feature learning, knowledge reasoning, and structured learning with observable performance gain for the task of aspect-based opinion extraction.

In this paper, we study the leading eigenvector problem in a statistically distributed setting and propose a communication-efficient algorithm based on Riemannian optimization, which trades local computation for global communication. Theoretical analysis shows that the proposed algorithm linearly converges to the centralized empirical risk minimization solution regarding the number of communication rounds. When the number of data points in local machines is sufficiently large, the proposed algorithm achieves a significant reduction of communication cost over existing distributed PCA algorithms. Superior performance in terms of communication cost of the proposed algorithm is verified on real-world and synthetic datasets.

In this paper, we propose a deep reinforcement learning (DRL) based framework to efficiently perform runtime channel pruning on convolutional neural networks (CNNs). Our DRL-based framework aims to learn a pruning strategy to determine how many and which channels to be pruned in each convolutional layer, depending on each individual input instance at runtime. Unlike existing runtime pruning methods which require to store all channels parameters for inference, our framework can reduce parameters storage consumption by introducing a static pruning component. Comparison experimental results with existing runtime and static pruning methods on state-of-the-art CNNs demonstrate that our proposed framework is able to provide a tradeoff between dynamic flexibility and storage efficiency in runtime channel pruning.

Heterogeneous domain adaptation (HDA) aims to solve the learning problems where the source- and the target-domain data are represented by heterogeneous types of features. The existing HDA approaches based on matrix completion or matrix factorization have proven to be effective to capture shareable information between heterogeneous domains. However, there are two limitations in the existing methods. First, a large number of corresponding data instances between the source domain and the target domain are required to bridge the gap between different domains for performing matrix completion. These corresponding data instances may be difficult to collect in real-world applications due to the limited size of data in the target domain. Second, most existing methods can only capture linear correlations between features and data instances while performing matrix completion for HDA. In this paper, we address these two issues by proposing a new matrix-factorization-based HDA method in a semisupervised manner, where only a few labeled data are required in the target domain without requiring any corresponding data instances between domains. Such labeled data are more practical to obtain compared with cross-domain corresponding data instances. Our proposed algorithm is based on matrix factorization in an approximated reproducing kernel Hilbert space (RKHS), where nonlinear correlations between features and data instances can be exploited to learn heterogeneous features for both the source and the target domains. Extensive experiments are conducted on cross-domain text classification and object recognition, and experimental results demonstrate the superiority of our proposed method compared with the state-of-the-art HDA approaches.

The advancement of deep models poses great challenges to real-world deployment because of the limited computational ability and storage space on edge devices. To solve this problem, existing works have made progress to prune or quantize deep models. However, most existing methods rely heavily on a supervised training process to achieve satisfactory performance, acquiring large amount of labeled training data, which may not be practical for real deployment. In this paper, we propose a novel layer-wise quantization method for deep neural networks, which only requires limited training data (1% of original dataset). Specifically, we formulate parameters quantization for each layer as a discrete optimization problem, and solve it using Alternative Direction Method of Multipliers (ADMM), which gives an efficient closed-form solution. We prove that the final performance drop after quantization is bounded by a linear combination of the reconstructed errors caused at each layer. Based on the proved theorem, we propose an algorithm to quantize a deep neural network layer by layer with an additional weights update step to minimize the final error. Extensive experiments on benchmark deep models are conducted to demonstrate the effectiveness of our proposed method using 1% of CIFAR10 and ImageNet datasets. Codes are available in: https://github.com/csyhhu/L-DNQ

Heterogeneous Transfer Learning (HTL) aims to solve transfer learning problems where a source domain and a target domain are of heterogeneous types of features. Most existing HTL approaches either explicitly learn feature mappings between the heterogeneous domains or implicitly reconstruct heterogeneous cross-domain features based on matrix completion techniques. In this paper, we propose a new HTL method based on a deep matrix completion framework, where kernel embedding of distributions is trained in an adversarial manner for learning heterogeneous features across domains. We conduct extensive experiments on two different vision tasks to demonstrate the effectiveness of our proposed method compared with a number of baseline methods.

Supervised learning methods have been widely applied to activity recognition. The prevalent success of existing methods, however, has two crucial prerequisites: proper feature extraction and sufficient labeled training data. The former is important to differentiate activities, while the latter is crucial to build a precise learning model. These two prerequisites have become bottlenecks to make existing methods more practical. Most existing feature extraction methods highly depend on domain knowledge, while labeled data requires intensive human annotation effort. Therefore, in this paper, we propose a novel method, named Distribution-based Semi-Supervised Learning, to tackle the aforementioned limitations. The proposed method is capable of automatically extracting powerful features with no domain knowledge required, meanwhile, alleviating the heavy annotation effort through semi-supervised learning. Specifically, we treat data stream of sensor readings received in a period as a distribution, and map all training distributions, including labeled and unlabeled, into a reproducing kernel Hilbert space (RKHS) using the kernel mean embedding technique. The RKHS is further altered by exploiting the underlying geometry structure of the unlabeled distributions. Finally, in the altered RKHS, a classifier is trained with the labeled distributions. We conduct extensive experiments on three public datasets to verify the effectiveness of our method compared with state-of-the-art baselines.

In fine-grained opinion mining, aspect and opinion terms extraction has become a fundamental task that provides key information for user-generated texts. Despite its importance, a lack of annotated resources in many domains impede the ability to train a precise model. Very few attempts have applied unsupervised domain adaptation methods to transfer fine-grained knowledge (in the word level) from some labeled source domain(s) to any unlabeled target domain. Existing methods depend on the construction of “pivot” knowledge, e.g., common opinion terms or syntactic relations between aspect and opinion words. In this work, we propose an interactive memory network that consists of local and global memory units. The model could exploit both local and global memory interactions to capture intra-correlations among aspect words or opinion words themselves, as well as the interconnections between aspect and opinion words. The source space and the target space are aligned through these domaininvariant interactions by incorporating an auxiliary task and domain adversarial networks. The proposed model does not require any external resources and demonstrates promising results on 3 benchmark datasets.

Recent years have witnessed the success of learning to hash in fast large-scale image retrieval. As deep learning has shown its superior performance on many computer vision applications, recent designs of learning-based hashing models have been moving from shallow ones to deep architectures. However, based on our analysis, we find that gradient descent based algorithms used in deep hashing models would potentially cause hash codes of a pair of training instances to be updated towards the directions of each other simultaneously during optimization. In the worst case, the paired hash codes switch their directions after update, and consequently, their corresponding distance in the Hamming space remain unchanged. This makes the overall learning process highly inefficient. To address this issue, we propose a new deep hashing model integrated with a novel gradient attention mechanism. Extensive experimental results on three benchmark datasets show that our proposed algorithm is able to accelerate the learning process and obtain competitive retrieval performance compared with state-of-the-art deep hashing models.

The advancement of deep models poses great challenges to real-world deployment because of the limited computational ability and storage space on edge devices. To solve this problem, existing works have made progress to compress deep models by pruning or quantization. However, most existing methods rely on a large amount of training data and a pre-trained model in the same domain. When only limited in-domain training data is available, these methods fail to perform well. This prompts the idea of transferring knowledge from a resource-rich source domain to a target domain with limited data to perform model compression. In this paper, we propose a method to perform cross-domain pruning by cooperatively training in both domains: taking advantage of data and a pre-trained model from the source domain to assist pruning in the target domain. Specifically, source and target pruned models are trained simultaneously and interactively, with source information transferred through the construction of a cooperative pruning mask. Our method significantly improves pruning quality in the target domain, and shed light to model compression in the cross-domain setting.

Feature-engineering-based machine learning models and deep learning models have been explored for wearable-sensor-based human activity recognition. For both types of methods, one crucial research issue is how to extract proper features from the partitioned segments of multivariate sensor readings. Existing methods have different drawbacks: 1) feature-engineering-based methods are able to extract meaningful features, such as statistical or structural information underlying the segments, but usually require manual designs of features for different applications, which is time consuming, and 2) deep learning models are able to learn temporal and/or spatial features from the sensor data automatically, but fail to capture statistical information. In this paper, we propose a novel deep learning model to automatically learn meaningful features including statistical features, temporal features and spatial correlation features for activity recognition in a unified framework. Extensive experiments are conducted on four datasets to demonstrate the effectiveness of our proposed method compared with state-of-the-art baselines.

Tremendous amount of parameters make deep neural networks impractical to be deployed for edge-device-based real-world applications due to the limit of computational power and storage space. Existing studies have made progress on learning quantized deep models to reduce model size and energy consumption, i.e. converting full-precision weights ($r$'s) into discrete values ($q$'s) in a supervised training manner. However, the training process for quantization is non-differentiable, which leads to either infinite or zero gradients ($g_r$) w.r.t. $r$. To address this problem, most training-based quantization methods use the gradient w.r.t. $q$ ($g_q$) with clipping to approximate $g_r$ by Straight-Through-Estimator (STE) or manually design their computation. However, these methods only heuristically make training-based quantization applicable, without further analysis on how the approximated gradients can assist training of a quantized network. In this paper, we propose to learn $g_r$ by a neural network. Specifically, a meta network is trained using $g_q$ and $r$ as inputs, and outputs $g_r$ for subsequent weight updates. The meta network is updated together with the original quantized network. Our proposed method alleviates the problem of non-differentiability, and can be trained in an end-to-end manner. Extensive experiments are conducted with CIFAR10/100 and ImageNet on various deep networks to demonstrate the advantage of our proposed method in terms of a faster convergence rate and better performance. Codes are released at: \texttt{https://github.com/csyhhu/MetaQuant}

In fine-grained opinion mining, extracting aspect terms (a.k.a. opinion targets) and opinion terms (a.k.a. opinion expressions) from user-generated texts is the most fundamental task in order to generate structured opinion summarization. Existing studies have shown that the syntactic relations between aspect and opinion words play an important role for aspect and opinion terms extraction. However, most of the works either relied on predefined rules or separated relation mining with feature learning. Moreover, these works only focused on single-domain extraction, which failed to adapt well to other domains of interest where only unlabeled data are available. In real-world scenarios, annotated resources are extremely scarce for many domains, motivating knowledge transfer strategies from labeled source domain(s) to any unlabeled target domain. We observe that syntactic relations among target words to be extracted are not only crucial for single-domain extraction, but also serve as invariant “pivot” information to bridge the gap between different domains. In this article, we explore the constructions of recursive neural networks based on the dependency tree of each sentence for associating syntactic structure with feature learning. Furthermore, we construct transferable recursive neural networks to automatically learn the domain-invariant fine-grained interactions among aspect words and opinion words. The transferability is built on an auxiliary task and a conditional domain adversarial network to reduce domain distribution difference in the hidden spaces effectively in word level through syntactic relations. Specifically, the auxiliary task builds structural correspondences across domains by predicting the dependency relation for each path of the dependency tree in the recursive neural network. The conditional domain adversarial network helps to learn domain-invariant hidden representation for each word conditioned on the syntactic structure. In the end, we integrate the recursive neural network with a sequence labeling classifier on top that models contextual influence in the final predictions. Extensive experiments and analysis are conducted to demonstrate the effectiveness of the proposed model and each component on three benchmark data sets.

A crucial issue in heterogeneous domain adaptation (HDA) is the ability to learn a feature mapping between different types of features across domains. Inspired by language translation, a word translated from one language corresponds to only a few words in another language, we present an efficient method named Sparse Heterogeneous Feature Representation (SHFR) in this paper for multi-class HDA to learn a sparse feature transformation between domains with multiple classes. Specifically, we formulate the problem of learning the feature transformation as a compressed sensing problem by building multiple binary classifiers in the target domain as various measurement sensors, which are decomposed from the target multi-class classification problem. We show that the estimation error of the learned transformation decreases with the increasing number of binary classifiers. In other words, for adaptation across heterogeneous domains to be successful, it is necessary to construct a sufficient number of incoherent binary classifiers from the original multi-class classification problem. To achieve this, we propose to apply the error correcting output correcting (ECOC) scheme to generate incoherent classifiers. To speed up the learning of the feature transformation across domains, we apply an efficient batch-mode algorithm to solve the resultant nonnegative sparse recovery problem. Theoretically, we present a generalization error bound of our proposed HDA method under a multi-class setting. Lastly, we conduct extensive experiments on both synthetic and real-world datasets to demonstrate the superiority of our proposed method over existing state-of-the-art HDA methods in terms of prediction accuracy and training efficiency.

The ever-expanding scale of WiFi deployments in metropolitan areas has made accurate GPS-free outdoor localization possible by relying solely on the WiFi infrastructure. Nevertheless, neither academic researches nor existing industrial practices seem to provide a satisfactory solution or implementation. In this paper, we propose WOLoc (WiFi-only Outdoor Localization) as a learning-based outdoor localization solution using only WiFi hotspots labeled by crowdsensing. On one hand, we do not take these labels as fingerprints as it is almost impossible to extend indoor localization mechanisms by fingerprinting metropolitan areas. On the other hand, we avoid the over-simplified local synthesis methods (e.g., centroid) that significantly lose the information contained in the labels. Instead, WOLoc adopts a semi-supervised manifold learning approach that accommodates all the labeled and unlabeled data for a given area, and the output concerning the unlabeled part will become the estimated locations for both unknown users and unknown WiFi hotspots. Moreover, WOLoc applies text mining techniques to analyze the SSIDs of hotspots, so as to derive more accurate input to its manifold learning. We conduct extensive experiments in several outdoor areas, and the results have strongly indicated the efficacy of our solution in achieving a meter-level localization accuracy.

Sensor-based activity recognition aims to predict users' activities from multi-dimensional streams of various sensor readings received from ubiquitous sensors. To use machine learning techniques for sensor-based activity recognition, previous approaches focused on composing a feature vector to represent sensor-reading streams received within a period of various lengths. With the constructed feature vectors, e.g., using predefined orders of moments in statistics, and their corresponding labels of activities, standard classification algorithms can be applied to train a predictive model, which will be used to make predictions online. However, we argue that in this way some important information, e.g., statistical information captured by higher-order moments, may be discarded when constructing features. Therefore, in this paper, we propose a new method, denoted by SMMAR, based on learning from distributions for sensor-based activity recognition. Specifically, we consider sensor readings received within a period as a sample, which can be represented by a feature vector of infinite dimensions in a Reproducing Kernel Hilbert Space (RKHS) using kernel embedding techniques. We then train a classifier in the RKHS. To scale-up the proposed method, we further offer an accelerated version by utilizing an explicit feature map instead of using a kernel function. We conduct experiments on four benchmark datasets to verify the effectiveness and scalability of our proposed method.

Multi-task learning (MTL) is a machine learning paradigm that improves the performance of each task by exploiting useful information contained in multiple related tasks. However, the relatedness of tasks can be exploited by attackers to launch data poisoning attacks, which has been demonstrated a big threat to single-task learning. In this paper, we provide the first study on the vulnerability of MTL. Specifically, we focus on multi-task relationship learning (MTRL) models, a popular subclass of MTL models where task relationships are quantized and are learned directly from training data. We formulate the problem of computing optimal poisoning attacks on MTRL as a bilevel program that is adaptive to arbitrary choice of target tasks and attacking tasks. We propose an efficient algorithm called PATOM for computing optimal attack strategies. PATOM leverages the optimality conditions of the subproblem of MTRL to compute the implicit gradients of the upper level objective function. Experimental results on real-world datasets show that MTRL models are very sensitive to poisoning attacks and the attacker can significantly degrade the performance of target tasks, by either directly poisoning the target tasks or indirectly poisoning the related tasks exploiting the task relatedness. We also found that the tasks being attacked are always strongly correlated, which provides a clue for defending against such attacks.

The iterative hard-thresholding algorithm (ISTA) is one of the most popular optimization solvers to achieve sparse codes. However, ISTA suffers from following problems: 1) ISTA employs non-adaptive updating strategy to learn the parameters on each dimension with a fixed learning rate. Such a strategy may lead to inferior performance due to the scarcity of diversity; 2) ISTA does not incorporate the historical information into the updating rules, and the historical information has been proven helpful to speed up the convergence. To address these challenging issues, we propose a novel formulation of ISTA (named as adaptive ISTA) by introducing a novel \textit{adaptive momentum vector}. To efficiently solve the proposed adaptive ISTA, we recast it as a recurrent neural network unit and show its connection with the well-known long short term memory (LSTM) model. With a new proposed unit, we present a neural network (termed SC2Net) to achieve sparse codes in an end-to-end manner. To the best of our knowledge, this is one of the first works to bridge the $\ell_1$-solver and LSTM, and may provide novel insights in understanding model-based optimization and LSTM. Extensive experiments show the effectiveness of our method on both unsupervised and supervised tasks.

Fine-grained opinion analysis aims to extract aspect and opinion terms from each sentence for opinion summarization.Supervised learning methods have proven to be effective for this task.However, in many domains, the lack of labeled data hinders the learning of a precise extraction model.In this case, unsupervised domain adaptation methods are desired to transfer knowledge from the source domain to any unlabeled target domain.In this paper, we develop a novel recursive neural network that could reduce domain shift effectively in word level through syntactic relations.We treat these relations as invariant "pivot information" across domains to build structural correspondences and generate an auxiliary task to predict the relation between any two adjacent words in the dependency tree.In the end, we demonstrate state-ofthe-art results on three benchmark datasets.

In this paper, we tackle the problem of domain generalization: how to learn a generalized feature representation for an "unseen" target domain by taking the advantage of multiple seen source-domain data. We present a novel framework based on adversarial autoencoders to learn a generalized latent feature representation across domains for domain generalization. To be specific, we extend adversarial autoencoders by imposing the Maximum Mean Discrepancy (MMD) measure to align the distributions among different domains, and matching the aligned distribution to an arbitrary prior distribution via adversarial feature learning. In this way, the learned feature representation is supposed to be universal to the seen source domains because of the MMD regularization, and is expected to generalize well on the target domain because of the introduction of the prior distribution. We proposed an algorithm to jointly train different components of our proposed framework. Extensive experiments on various vision tasks demonstrate that our proposed framework can learn better generalized features for the unseen target domain compared with state-of-the-art domain generalization methods.

In fine-grained opinion mining, the task of aspect extraction involves the identification of explicit product features in customer reviews. This task has been widely studied in some major languages, e.g., English, but was seldom addressed in other minor languages due to the lack of annotated corpus. To solve it, we develop a novel deep model to transfer knowledge from a source language with labeled training data to a target language without any annotations. Different from cross-lingual sentiment classification, aspect extraction across languages requires more fine-grained adaptation. To this end, we utilize transition-based mechanism that reads a word each time and forms a series of configurations that represent the status of the whole sentence. We represent each configuration as a continuous feature vector and align these representations from different languages into a shared space through an adversarial network. In addition, syntactic structures are also integrated into the deep model to achieve more syntactically-sensitive adaptations. The proposed method is end-to-end and achieves state-of-the-art performance on English, French and Spanish restaurant review datasets.

In deep reinforcement learning (RL) tasks, an efficient exploration mechanism should be able to encourage an agent to take actions that lead to less frequent states which may yield higher accumulative future return. However, both knowing about the future and evaluating the frequentness of states are non-trivial tasks, especially for deep RL domains, where a state is represented by high-dimensional image frames. In this paper, we propose a novel informed exploration framework for deep RL, where we build the capability for an RL agent to predict over the future transitions and evaluate the frequentness for the predicted future frames in a meaningful manner. To this end, we train a deep prediction model to predict future frames given a state-action pair, and a convolutional autoencoder model to hash over the seen frames. In addition, to utilize the counts derived from the seen frames to evaluate the frequentness for the predicted frames, we tackle the challenge of matching the predicted future frames and their corresponding seen frames at the latent feature level. In this way, we derive a reliable metric for evaluating the novelty of the future direction pointed by each action, and hence inform the agent to explore the least frequent one.

An autonomous framework to evaluate layout of a piping design in the form of piping and instrumentation diagram (P&ID) according to a set of standards of marine and offshore industry is proposed. The method starts with transforming a P&ID into a vector x in Rd. Transformation is done based on a concept introduced for piping known as Histogram of Connectivity. The proposed descriptor captures two essential properties of P&ID: attributes of each component and connectivity among the components. Next, linear support vector machine (SVM) is used to learn a classifier from existing compliant and noncompliant designs. Subsequently, the linear classifier can be used to check if an unseen design complies with the standards. In addition, to enable follow up on noncompliant design including correction or modification, a method to analyze the reason of noncompliance prediction by the learned SVM model is introduced. The method has demonstrated encouraging performance in two challenging data sets of designs created with advice from experienced engineers in the industry, based on International Convention for the Prevention of Pollution from Ships (MARPOL) and Rules for Classification of Ships of Lloyd's Register.

Transfer learning has gained a lot of attention and interest in the past decade. One crucial research issue in transfer learning is how to find a good representation for instances of different domains such that the divergence between domains can be reduced with the new representation. Recently, deep learning has been proposed to learn more robust or higher-level features for transfer learning. In this article, we adapt the autoencoder technique to transfer learning and propose a supervised representation learning method based on double encoding-layer autoencoder. The proposed framework consists of two encoding layers: one for embedding and the other one for label encoding. In the embedding layer, the distribution distance of the embedded instances between the source and target domains is minimized in terms of KL-Divergence. In the label encoding layer, label information of the source domain is encoded using a softmax regression model. Moreover, to empirically explore why the proposed framework can work well for transfer learning, we propose a new effective measure based on autoencoder to compute the distribution distance between different domains. Experimental results show that the proposed new measure can better reflect the degree of transfer difficulty and has stronger correlation with the performance from supervised learning algorithms (e.g., Logistic Regression), compared with previous ones, such as KL-Divergence and Maximum Mean Discrepancy. Therefore, in our model, we have incorporated two distribution distance measures to minimize the difference between source and target domains in the embedding representations. Extensive experiments conducted on three real-world image datasets and one text data demonstrate the effectiveness of our proposed method compared with several state-of-the-art baseline methods.

In this paper, we study a novel transfer learning problem termed Distant Domain Transfer Learning (DDTL). Different from existing transfer learning problems which assume that there is a close relation between the source domain and the target domain, in the DDTL problem, the target domain can be totally different from the source domain. For example, the source domain classifies face images but the target domain distinguishes plane images. Inspired by the cognitive processof human where two seemingly unrelated concepts can be connected by learning intermediate concepts gradually, we propose a Selective Learning Algorithm (SLA) to solve the DDTL problem with supervised autoencoder or supervised convolutional autoencoder as a base model for handling different types of inputs. Intuitively, the SLA algorithm selects usefully unlabeled data gradually from intermediate domains as a bridge to break the large distribution gap for transferring knowledge between two distant domains. Empirical studies on image classification problems demonstrate the effectiveness of the proposed algorithm, and on some tasks the improvement in terms of the classification accuracy is up to 17% over “non-transfer” methods.

The task of aspect and opinion terms co-extraction aims to explicitly extract aspect terms describing features of an entity and opinion terms expressing emotions from user-generated texts. To achieve this task, one effective approach is to exploit relations between aspect terms and opinion terms by parsing syntactic structure for each sentence. However, this approach requires expensive effort for parsing and highly depends on the quality of the parsing results. In this paper, we offer a novel deep learning model, named coupled multi-layer attentions. The proposed model provides an end-to-end solution and does not require any parsers or other linguistic resources for preprocessing. Specifically, the proposed model is a multi-layer attention network, where each layer consists of a couple of attentions with tensor operators. One attention is for extracting aspect terms, while the other is for extracting opinion terms. They are learned interactively to dually propagate information between aspect terms and opinion terms. Through multiple layers, the model can further exploit indirect relations between terms for more precise information extraction. Experimental results on three benchmark datasets in SemEval Challenge 2014 and 2015 show that our model achieves state-of-the-art performances compared with several baselines.

The process for transferring knowledge of multiple reinforcement learning policies into a single multi-task policy via distillation technique is known as policy distillation. When policy distillation is under a deep reinforcement learning setting, due to the giant parameter size and the huge state space for each task domain, it requires extensive computational efforts to train the multi-task policy network. In this paper, we propose a new policy distillation architecture for deep reinforcement learning, where we assume that each task uses its task-specific high-level convolutional features as the inputs to the multi-task policy network. Furthermore, we propose a new sampling framework termed hierarchical prioritized experience replay to selectively choose experiences from the replay memories of each task domain to perform learning on the network. With the above two attempts, we aim to accelerate the learning of the multi-task policy network while guaranteeing a good performance. We use Atari 2600 games as testing environment to demonstrate the efficiency and effectiveness of our proposed solution for policy distillation

In multi-task learning (MTL), tasks are learned jointly so that information among related tasks is shared and utilized to help improve generalization for each individual task. A major challenge in MTL is how to selectively choose what to share among tasks. Ideally, only related tasks should share information with each other. In this paper, we propose a new MTL method that can adaptively group correlated tasks into clusters and share information among the correlated tasks only. Our method is based on the assumption that each task parameter is a linear combination of other tasks' and the coefficients of the linear combination are active only if there is relatedness between the two tasks. Through introducing trace Lasso penalty on these coefficients, our method is able to adaptively select the subset of coefficients with respect to the tasks that are correlated to the task. Our model frees the process of determining task clustering structure as used in the literature. Efficient optimization methods based on alternating direction method of multipliers (ADMM) is developed to solve the problem. Experimental results on both synthetic and real-world datasets demonstrate the effectiveness of our method in terms of clustering related tasks and generalization performance.

Given the ever-expanding scale of WiFi deployments in metropolitan areas, we have reached the point where accurate GPS-free outdoor localization becomes possible by relying solely on the WiFi infrastructure. Nevertheless, the existing industrial practices do not seem to have the right implementation to achieve an adequate accuracy, while the academic researches that are mostly attracted by indoor localization have largely neglected this outdoor aspect. In this paper, we propose WOLoc (WiFi-only Outdoor Localization) as a solution that offers meter-level accuracy, by holistically treating the large number of WiFi hotspot labels gather by crowdsensing. On one hand, we do not take these labels as fingerprints as it is almost impossible to extend indoor localization mechanisms by fingerprinting metropolitan areas. On the other hand, we avoid the over-simplified local synthesis methods (e.g., centroid) that significantly lose the information contained in the labels. Instead, we accommodate all the labeled and unlabeled data for a given area using a semi-supervised manifold learning technique, and the output concerning the unlabeled part will become the estimated locations for both users and WiFi hotspots. We conduct extensive experiments with WOLoc in several outdoor areas, and the results have strongly indicated the efficacy of our solution.

Multi-task learning aims to learn multiple tasks jointly by exploiting their relatedness to improve the generalization performance for each task. Traditionally, to perform multi-task learning, one needs to centralize data from all the tasks to a single machine. However, in many real-world applications, data of different tasks may be geo-distributed over different local machines. Due to heavy communication caused by transmitting the data and the issue of data privacy and security, it is impossible to send data of different task to a master machine to perform multi-task learning. Therefore, in this paper, we propose a distributed multi-task learning framework that simultaneously learns predictive models for each task as well as task relationships between tasks alternatingly in the parameter server paradigm. In our framework, we first offer a general dual form for a family of regularized multi-task relationship learning methods. Subsequently, we propose a communication-efficient primal-dual distributed optimization algorithm to solve the dual problem by carefully designing local subproblems to make the dual problem decomposable. Moreover, we provide a theoretical convergence analysis for the proposed algorithm, which is specific for distributed multi-task relationship learning. We conduct extensive experiments on both synthetic and real-world datasets to evaluate our proposed framework in terms of effectiveness and convergence.

How to develop slim and accurate deep neural networks has become crucial for real- world applications, especially for those employed in embedded systems. Though previous work along this research line has shown some promising results, most existing methods either fail to significantly compress a well-trained deep network or require a heavy retraining process for the pruned deep network to re-boost its prediction performance. In this paper, we propose a new layer-wise pruning method for deep neural networks. In our proposed method, parameters of each individual layer are pruned independently based on second order derivatives of a layer-wise error function with respect to the corresponding parameters. We prove that the final prediction performance drop after pruning is bounded by a linear combination of the reconstructed errors caused at each layer. By controlling layer-wise errors properly, one only needs to perform a light retraining process on the pruned network to resume its original prediction performance. We conduct extensive experiments on benchmark datasets to demonstrate the effectiveness of our pruning method compared with several state-of-the-art baseline methods. Codes of our work are released at: https://github.com/csyhhu/L-OBS.

As a key component of building management and security, occupancy inference through smart sensing has attracted a lot of research attentions for nearly two decades. Nevertheless, existing solutions mostly rely on either pre-deployed infrastructures or user device participation, thus hampering their wide adoption. This paper presents CeilingSee, a dedicated occupancy inference system free of heavy infrastructure deployments and user involvements. Building upon existing LED lighting systems, CeilingSee converts part of the ceiling-mounted LED luminaires to act as sensors, sensing the variances in diffuse reflection caused by occupants. In realizing CeilingSee, we first re-design the LED driver to leverage LED's photoelectric effect so as to transform a light emitter to a light sensor. In order to produce accurate occupancy inference, we then engineer efficient learning algorithms to fuse sensing information gathered by multiple LED luminaires. We build a testbed covering a 30m2office area; extensive experiments show that CeilingSee is able to achieve very high accuracy in occupancy inference.

Transfer learning has been proven to be effective for the problems where training data from a source domain and test data from a target domain are drawn from different distributions. To reduce the distribution divergence between the source domain and the target domain, many previous studies have been focused on designing and optimizing objective functions with the Euclidean distance to measure dissimilarity between instances. However, in some real-world applications, the Euclidean distance may be inappropriate to capture the intrinsic similarity or dissimilarity between instances. To deal with this issue, in this paper, we propose a metric transfer learning framework (MTLF) to encode metric learning in transfer learning. In MTLF, instance weights are learned and exploited to bridge the distributions of different domains, while Mahalanobis distance is learned simultaneously to maximize the intra-class distances and minimize the inter-class distances for the target domain. Unlike previous work where instance weights and Mahalanobis distance are trained in a pipelined framework that potentially leads to error propagation across different components, MTLF attempts to learn instance weights and a Mahalanobis distance in a parallel framework to make knowledge transfer across domains more effective. Furthermore, we develop general solutions to both classification and regression problems on top of MTLF, respectively. We conduct extensive experiments on several real-world datasets on object recognition, handwriting recognition, and WiFi location to verify the effectiveness of MTLF compared with a number of state-of-the-art methods.

Most existing heterogeneous transfer learning (HTL) methods for cross-language text classification rely on sufficient cross-domain instance correspondences to learn a mapping across heterogeneous feature spaces, and assume that such correspondences are given in advance. However, in practice, correspondences between domains are usually unknown. In this case, extensively manual efforts are required to establish accurate correspondences across multilingual documents based on their content and meta-information. In this paper, we present a general framework to integrate active learning to construct correspondences between heterogeneous domains for HTL, namely HTL through active correspondences construction (HTLA). Based on this framework, we develop a new HTL method. On top of the new HTL method, we further propose a strategy to actively construct correspondences between domains. Extensive experiments are conducted on various multilingual text classification tasks to verify the effectiveness of HTLA.

In the past decade, there have been a large number of transfer learning algorithms proposed for various real-world applications. However, most of them are vulnerable to negative transfer since their performance is even worse than traditional supervised models. Aiming at more robust transfer learning models, we propose an ENsemble framework of anCHOR adapters (ENCHOR for short), in which an anchor adapter adapts the features of instances based on their similarities to a specific anchor (i.e., a selected instance). Specifically, the more similar to the anchor instance, the higher degree of the original feature of an instance remains unchanged in the adapted representation, and vice versa. This adapted representation for the data actually expresses the local structure around the corresponding anchor, and then any transfer learning method can be applied to this adapted representation for a prediction model, which focuses more on the neighborhood of the anchor. Next, based on multiple anchors, multiple anchor adapters can be built and combined into an ensemble for final output. Additionally, we develop an effective measure to select the anchors for ensemble building to achieve further performance improvement. Extensive experiments on hundreds of text classification tasks are conducted to demonstrate the effectiveness of ENCHOR. The results show that: when traditional supervised models perform poorly, ENCHOR (based on only 8 selected anchors) achieves $6%-13%$ increase in terms of average accuracy compared with the state-of-the-art methods, and it greatly alleviates negative transfer.

In aspect-based sentiment analysis, extracting aspect terms along with the opinions being expressed from user-generated content is one of the most important subtasks.Previous studies have shown that exploiting connections between aspect and opinion terms is promising for this task.In this paper, we propose a novel joint model that integrates recursive neural networks and conditional random fields into a unified framework for explicit aspect and opinion terms co-extraction.The proposed model learns high-level discriminative features and double propagates information between aspect and opinion terms, simultaneously.Moreover, it is flexible to incorporate hand-crafted features into the proposed model to further boost its information extraction performance.Experimental results on the dataset from SemEval Challenge 2014 task 4 show the superiority of our proposed model over several baseline methods as well as the winning systems of the challenge.

Most software defect prediction approaches are trained and applied on data from the same project. However, often a new project does not have enough training data. Cross-project defect prediction, which uses data from other projects to predict defects in a particular project, provides a new perspective to defect prediction. In this work, we propose a HYbrid moDel Reconstruction Approach (HYDRA) for cross-project defect prediction, which includes two phases: genetic algorithm (GA) phase and ensemble learning (EL) phase. These two phases create a massive composition of classifiers. To examine the benefits of HYDRA, we perform experiments on 29 datasets from the PROMISE repository which contains a total of 11,196 instances (i.e., Java classes) labeled as defective or clean. We experiment with logistic regression as the underlying classification algorithm of HYDRA. We compare our approach with the most recently proposed cross-project defect prediction approaches: TCA+ by Nam et al., Peters filter by Peters et al., GP by Liu et al., MO by Canfora et al., and CODEP by Panichella et al. Our results show that HYDRA achieves an average F1-score of 0.544. On average, across the 29 datasets, these results correspond to an improvement in the F1-scores of 26.22 , 34.99, 47.43, 28.61, and 30.14 percent over TCA+, Peters filter, GP, MO, and CODEP, respectively. In addition, HYDRA on average can discover 33 percent of all bugs if developers inspect the top 20 percent lines of code, which improves the best baseline approach (TCA+) by 44.41 percent. We also find that HYDRA improves the F1-score of Zero-R which predict all the instances to be defective by 5.42 percent, but improves Zero-R by 58.65 percent when inspecting the top 20 percent lines of code. In practice, Zero-R can be hard to use since it simply predicts all of the instances to be defective, and thus developers have to inspect all of the instances to find the defective ones. Moreover, we notice the improvement of HYDRA over other baseline approaches in terms of F1-score and when inspecting the top 20 percent lines of code are substantial, and in most cases the improvements are significant and have large effect sizes across the 29 datasets.

Multi-task Learning (MTL) aims to learn multiple related tasks simultaneously instead of separately to improve generalization performance of each task. Most existing MTL methods assumed that the multiple tasks to be learned have the same feature representation. However, this assumption may not hold for many real-world applications. In this paper, we study the problem of MTL with heterogeneous features for each task. To address this problem, we first construct an integrated graph of a set of bipartite graphs to build a connection among different tasks. We then propose a multi-task nonnegative matrix factorization (MTNMF) method to learn a common semantic feature space underlying different heterogeneous feature spaces of each task. Finally, based on the common semantic features and original heterogeneous features, we model the heterogenous MTL problem as a multi-task multi-view learning (MTMVL) problem. In this way, a number of existing MTMVL methods can be applied to solve the problem effectively. Extensive experiments on three real-world problems demonstrate the effectiveness of our proposed method.

Transfer learning uses relevant auxiliary data to help the learning task in a target domain where labeled data is usually insufficient to train an accurate model. Given appropriate auxiliary data, researchers have proposed many transfer learning models. How to find such auxiliary data, however, is of little research so far. In this paper, we focus on the problem of auxiliary data retrieval, and propose a transfer learning framework that effectively selects helpful auxiliary data from an open knowledge space (e.g. the World Wide Web). Because there is no need of manually selecting auxiliary data for different target domain tasks, we call our framework Source Free Transfer Learning (SFTL). For each target domain task, SFTL framework iteratively queries for the helpful auxiliary data based on the learned model and then updates the model using the retrieved auxiliary data. We highlight the automatic constructions of queries and the robustness of the SFTL framework. Our experiments on 20NewsGroup dataset and a Google search snippets dataset suggest that the framework is capable of achieving comparable performance to those state-of-the-art methods with dedicated selections of auxiliary data.

Most previous heterogeneous transfer learning methods learn a cross-domain feature mapping between heterogeneous feature spaces based on a few cross-domain instance-correspondences, and these corresponding instances are assumed to be representative in the source and target domains respectively. However, in many real-world scenarios, this assumption may not hold. As a result, the constructed feature mapping may not be precisely due to the bias issue of the correspondences in the target or (and) source domain(s). In this case, a classifier trained on the labeled transformed-source-domain data may not be useful for the target domain. In this paper, we present a new transfer learning framework called Hybrid Heterogeneous Transfer Learning (HHTL), which allows the corresponding instances across domains to be biased in either the source or target domain. Specifically, we propose a deep learning approach to learn a feature mapping between cross-domain heterogeneous features as well as a better feature representation for mapped data to reduce the bias issue caused by the cross-domain correspondences. Extensive experiments on several multilingual sentiment classification tasks verify the effectiveness of our proposed approach compared with some baseline methods.

In this paper, we present an efficient Multi-class Heterogeneous Domain Adaptation (HDA) method, where data from the source and target domains are represented by heterogeneous features with different dimensions. Specifically, we propose to reconstruct a sparse feature transformation matrix to map the features of multiple classes from the source domain to the target domain. We cast this learning task as a compressed sensing problem, where each classifier can be deemed as a measurement sensor. Based on compressive sensing theory, the estimation error of the transformation matrix decreases with the increasing number of classifiers. Therefore, to guarantee the reconstruction performance, we construct sufficiently many binary classifiers based on the error correcting output correcting. Extensive experiments are conducted on both toy data and three real-world HDA applications to verify the superiority of our proposed method over existing state-of-the-art HDA methods in terms of prediction accuracy.

In this paper we first exploit cash-tags ("$" followed by stocks' ticker symbols) in Twitter to build a stock network, where nodes are stocks connected by edges when two stocks co-occur frequently in tweets.We then employ a labeled topic model to jointly model both the tweets and the network structure to assign each node and each edge a topic respectively.This Semantic Stock Network (SSN) summarizes discussion topics about stocks and stock relations.We further show that social sentiment about stock (node) topics and stock relationship (edge) topics are predictive of each stock's market.For prediction, we propose to regress the topic-sentiment time-series and the stock's price time series.Experimental results demonstrate that topic sentiments from close neighbors are able to help improve the prediction of a stock markedly.

Low rank matrix recovery is a fundamental task in many real-world applications. The performance of existing methods, however, deteriorates significantly when applied to ill-conditioned or large-scale matrices. In this paper, we therefore propose an efficient method, called Riemannian Pursuit (RP), that aims to address these two problems simultaneously. Our method consists of a sequence of fixed-rank optimization problems. Each subproblem, solved by a nonlinear Riemannian conjugate gradient method, aims to correct the solution in the most important subspace of increasing size. Theoretically, RP converges linearly under mild conditions and experimental results show that it substantially outperforms existing methods when applied to large-scale and ill-conditioned matrices.

Domain transfer learning, which learns a target classifier using labeled data from a different distribution, has shown promising value in knowledge discovery yet still been a challenging problem. Most previous works designed adaptive classifiers by exploring two learning strategies independently: distribution adaptation and label propagation. In this paper, we propose a novel transfer learning framework, referred to as Adaptation Regularization based Transfer Learning (ARTL), to model them in a unified way based on the structural risk minimization principle and the regularization theory. Specifically, ARTL learns the adaptive classifier by simultaneously optimizing the structural risk functional, the joint distribution matching between domains, and the manifold consistency underlying marginal distribution. Based on the framework, we propose two novel methods using Regularized Least Squares (RLS) and Support Vector Machines (SVMs), respectively, and use the Representer theorem in reproducing kernel Hilbert space to derive corresponding solutions. Comprehensive experiments verify that ARTL can significantly outperform state-of-the-art learning methods on several public text and image datasets.

In real-life problems, the following semi-supervised domain adaptation scenario is often encountered: we have full access to some source data, which is usually very large; the target data distribution is under certain unknown transformation of the source data distribution; meanwhile, only a small fraction of the target instances come with labels. The goal is to learn a prediction model by incorporating information from the source domain that is able to generalize well on the target test instances. We consider an explicit form of transformation functions and especially linear transformations that maps examples from the source to the target domain, and we argue that by proper preprocessing of the data from both source and target domains, the feasible transformation functions can be characterized by a set of rotation matrices. This naturally leads to an optimization formulation under the special orthogonal group constraints. We present an iterative coordinate descent solver that is able to jointly learn the transformation as well as the model parameters, while the geodesic update ensures the manifold constraints are always satisfied. Our framework is sufficiently general to work with a variety of loss functions and prediction problems. Empirical evaluations on synthetic and real-world experiments demonstrate the competitive performance of our method with respect to the state-of-the-art.

Recommender systems, especially the newly launched ones, have to deal with the data-sparsity issue, where little existing rating information is available. Recently, transfer learning has been proposed to address this problem by leveraging the knowledge from related recommender systems where rich collaborative data are available. However, most previous transfer learning models assume that entity-correspondences across different systems are given as input, which means that for any entity (e.g., a user or an item) in a target system, its corresponding entity in a source system is known. This assumption can hardly be satisfied in real-world scenarios where entity-correspondences across systems are usually unknown, and the cost of identifying them can be expensive. For example, it is extremely difficult to identify whether a user A from Facebook and a user B from Twitter are the same person. In this paper, we propose a framework to construct entity correspondence with limited budget by using active learning to facilitate knowledge transfer across recommender systems. Specifically, for the purpose of maximizing knowledge transfer, we first iteratively select entities in the target system based on our proposed criterion to query their correspondences in the source system. We then plug the actively constructed entity-correspondence mapping into a general transferred collaborative-filtering model to improve recommendation quality. We perform extensive experiments on real world datasets to verify the effectiveness of our proposed framework for this cross-system recommendation problem.

In this paper, we propose a new framework for opinion summarization based on sentence selection. Our goal is to assist users to get helpful opinion suggestions from reviews by only reading a short summary with few informative sentences, where the quality of summary is evaluated in terms of both aspect coverage and viewpoints preservation. More specifically, we formulate the informative-sentence selection problem in opinion summarization as a community-leader detection problem, where a community consists of a cluster of sentences towards the same aspect of an entity. The detected leaders of the communities can be considered as the most informative sentences of the corresponding aspect, while informativeness of a sentence is defined by its informativeness within both its community and the document it belongs to. Review data from six product domains from Amazon.com are used to verify the effectiveness of our method for opinion summarization.

The study of users' social behaviors has gained much research attention since the advent of various social media such as Facebook, Renren and Twitter. A major kind of applications is to predict a user's future activities based on his/her historical social behaviors. In this paper, we focus on a fundamental task: to predict a user's future activity levels in a social network, e.g. weekly activeness, active or inactive. This problem is closely related to Social Customer Relationship Management (Social CRM). Compared to traditional CRM, the three properties: user diversity, social influence, and dynamic nature of social networks, raise new challenges and opportunities to Social CRM. Firstly, the user diversity property implies that a global predictive model may not be precise for all users. On the other hand, historical data of individual users are too sparse to build precisely personalized models. Secondly, the social influence property suggests that relationships between users can be embedded to further boost prediction results on individual users. Finally, the dynamical nature of social networks means that users' behaviors may keep changing over time. To address these challenges, we develop a personalized and social regularized time-decay model for user activity level prediction. Experiments on the social media Renren validate the effectiveness of our proposed model compared with some baselines including traditional supervised learning methods and node classification methods in social networks.

Many software defect prediction approaches have been proposed and most are effective in within-project prediction settings. However, for new projects or projects with limited training data, it is desirable to learn a prediction model by using sufficient training data from existing source projects and then apply the model to some target projects (cross-project defect prediction). Unfortunately, the performance of cross-project defect prediction is generally poor, largely because of feature distribution differences between the source and target projects. In this paper, we apply a state-of-the-art transfer learning approach, TCA, to make feature distributions in source and target projects similar. In addition, we propose a novel transfer defect learning approach, TCA+, by extending TCA. Our experimental results for eight open-source projects show that TCA+ significantly improves cross-project prediction performance.

Named Entity Recognition (NER) is a fundamental task in information extraction from unstructured text. Most previous machine-learning-based NER systems are domain-specific, which implies that they may only perform well on some specific domains (e.g., Newswire ) but tend to adapt poorly to other related but different domains (e.g., Weblog ). Recently, transfer learning techniques have been proposed to NER. However, most transfer learning approaches to NER are developed for binary classification, while NER is a multiclass classification problem in nature. Therefore, one has to first reduce the NER task to multiple binary classification tasks and solve them independently. In this article, we propose a new transfer learning method, named Transfer Joint Embedding (TJE), for cross-domain multiclass classification, which can fully exploit the relationships between classes (labels), and reduce domain difference in data distributions for transfer learning. More specifically, we aim to embed both labels (outputs) and high-dimensional features (inputs) from different domains (e.g., a source domain and a target domain) into a unified low-dimensional latent space, where 1) each label is represented by a prototype and the intrinsic relationships between labels can be measured by Euclidean distance; 2) the distance in data distributions between the source and target domains can be reduced; 3) the source domain labeled data are closer to their corresponding label-prototypes than others. After the latent space is learned, classification on the target domain data can be done with the simple nearest neighbor rule in the latent space. Furthermore, in order to scale up TJE, we propose an efficient algorithm based on stochastic gradient descent (SGD). Finally, we apply the proposed TJE method for NER across different domains on the ACE 2005 dataset, which is a benchmark in Natural Language Processing (NLP). Experimental results demonstrate the effectiveness of TJE and show that TJE can outperform state-of-the-art transfer learning approaches to NER.

Active learning has been proven to be effective in reducing labeling efforts for supervised learning. However, existing active learning work has mainly focused on training models for a single domain. In practical applications, it is common to simultaneously train classifiers for multiple domains. For example, some merchant web sites (like Amazon.com) may need a set of classifiers to predict the sentiment polarity of product reviews collected from various domains (e.g., electronics, books, shoes). Though different domains have their own unique features, they may share some common latent features. If we apply active learning on each domain separately, some data instances selected from different domains may contain duplicate knowledge due to the common features. Therefore, how to choose the data from multiple domains to label is crucial to further reducing the human labeling efforts in multi-domain learning. In this paper, we propose a novel multi-domain active learning framework to jointly select data instances from all domains with duplicate information considered. In our solution, a shared subspace is first learned to represent common latent features of different domains. By considering the common and the domain-specific features together, the model loss reduction induced by each data instance can be decomposed into a common part and a domain-specific part. In this way, the duplicate information across domains can be encoded into the common part of model loss reduction and taken into account when querying. We compare our method with the state-of-the-art active learning approaches on several text classification tasks: sentiment classification, newsgroup classification and email spam filtering. The experiment results show that our method reduces the human labeling efforts by 33.2%, 42.9% and 68.7% on the three tasks, respectively.

Recent years have witnessed the growing popularity of sensor and sensor-network technologies, supporting important practical applications. One of the fundamental issues is how to accurately locate a user with few labeled data in a wireless sensor network, where a major difficulty arises from the need to label large quantities of user location data, which in turn requires knowledge about the locations of signal transmitters or access points. To solve this problem, we have developed a novel machine learning-based approach that combines collaborative filtering with graph-based semi-supervised learning to learn both mobile users' locations and the locations of access points. Our framework exploits both labeled and unlabeled data from mobile devices and access points. In our two-phase solution, we first build a manifold-based model from a batch of labeled and unlabeled data in an offline training phase and then use a weighted k-nearest-neighbor method to localize a mobile client in an online localization phase. We extend the two-phase colocalization to an online and incremental model that can deal with labeled and unlabeled data that come sequentially and adapt to environmental changes. Finally, we embed an action model to the framework such that additional kinds of sensor signals can be utilized to further boost the performance of mobile tracking. Compared to other state-of-the-art systems, our framework has been shown to be more accurate while requiring less calibration effort in our experiments performed on three different testbeds.

Transfer learning as a new machine learning paradigm has gained increasing attention lately. In situations where the training data in a target domain are not sufficient to learn predictive models effectively, transfer learning leverages auxiliary source data from other related source domains for learning. While most of the existing works in this area only focused on using the source data with the same structure as the target data, in this paper, we push this boundary further by proposing a heterogeneous transfer learning framework for knowledge transfer between text and images. We observe that for a target-domain classification problem, some annotated images can be found on many social Web sites, which can serve as a bridge to transfer knowledge from the abundant text documents available over the Web. A key question is how to effectively transfer the knowledge in the source data even though the text can be arbitrarily found. Our solution is to enrich the representation of the target images with semantic concepts extracted from the auxiliary source data through a novel matrix factorization method. By using the latent semantic features generated by the auxiliary data, we are able to build a better integrated image classifier. We empirically demonstrate the effectiveness of our algorithm on the Caltech-256 image dataset.

Transfer learning addresses the problems that labeled training data are insufficient to produce a high-performance model. Typically, given a target learning task, most transfer learning approaches require to select one or more auxiliary tasks as sources by the designers. However, how to select the right source data to enable effective knowledge transfer automatically is still an unsolved problem, which limits the applicability of transfer learning. In this paper, we take one step ahead and propose a novel transfer learning framework, known as source-selection-free transfer learning (SSFTL), to free users from the need to select source domains. Instead of asking the users for source and target data pairs, as traditional transfer learning does, SS-FTL turns to some online information sources such as World Wide Web or the Wikipedia for help. The source data for transfer learning can be hidden somewhere within this large online information source, but the users do not know where they are. Based on the online information sources, we train a large number of classifiers. Then, given a target task, a bridge is built for labels of the potential source candidates and the target domain data in SSFTL via some large online social media with tag cloud as a label translator. An added advantage of SSFTL is that, unlike many previous transfer learning approaches, which are difficult to scale up to the Web scale, SSFTL is highly scalable and can offset much of the training work to offline stage. We demonstrate the effectiveness and efficiency of SS-FTL through extensive experiments on several realworld datasets in text classification. 1

Protein subcellular localization is concerned with predicting the location of a protein within a cell using computational methods. The location information can indicate key functionalities of proteins. Thus, accurate prediction of subcellular localizations of proteins can help the prediction of protein functions and genome annotations, as well as the identification of drug targets. Machine learning methods such as Support Vector Machines (SVMs) have been used in the past for the problem of protein subcellular localization, but have been shown to suffer from a lack of annotated training data in each species under study. To overcome this data sparsity problem, we observe that because some of the organisms may be related to each other, there may be some commonalities across different organisms that can be discovered and used to help boost the data in each localization task. In this paper, we formulate protein subcellular localization problem as one of multitask learning across different organisms. We adapt and compare two specializations of the multitask learning algorithms on 20 different organisms. Our experimental results show that multitask learning performs much better than the traditional single-task methods. Among the different multitask learning methods, we found that the multitask kernels and supertype kernels under multitask learning that share parameters perform slightly better than multitask learning by sharing latent features. The most significant improvement in terms of localization accuracy is about 25 percent. We find that if the organisms are very different or are remotely related from a biological point of view, then jointly training the multiple models cannot lead to significant improvement. However, if they are closely related biologically, the multitask learning can do much better than individual learning.

Domain adaptation allows knowledge from a source domain to be transferred to a different but related target domain. Intuitively, discovering a good feature representation across domains is crucial. In this paper, we first propose to find such a representation through a new learning method, transfer component analysis (TCA), for domain adaptation. TCA tries to learn some transfer components across domains in a reproducing kernel Hilbert space using maximum mean miscrepancy. In the subspace spanned by these transfer components, data properties are preserved and data distributions in different domains are close to each other. As a result, with the new representations in this subspace, we can apply standard machine learning methods to train classifiers or regression models in the source domain for use in the target domain. Furthermore, in order to uncover the knowledge hidden in the relations between the data labels from the source and target domains, we extend TCA in a semisupervised learning setting, which encodes label information into transfer components learning. We call this extension semisupervised TCA. The main contribution of our work is that we propose a novel dimensionality reduction framework for reducing the distance between domains in a latent space for domain adaptation. We propose both unsupervised and semisupervised feature extraction approaches, which can dramatically reduce the distance between domain distributions by projecting data onto the learned transfer components. Finally, our approach can handle large datasets and naturally lead to out-of-sample generalization. The effectiveness and efficiency of our approach are verified by experiments on five toy datasets and two real-world applications: cross-domain indoor WiFi localization and cross-domain text classification.

Transfer learning aims at reusing the knowledge in some source tasks to improve the learning of a target task. Many transfer learning methods assume that the source tasks and the target task be related, even though many tasks are not related in reality. However, when two tasks are unrelated, the knowledge extracted from a source task may not help, and even hurt, the performance of a target task. Thus, how to avoid negative transfer and then ensure a "safe transfer" of knowledge is crucial in transfer learning. In this paper, we propose an Adaptive Transfer learning algorithm based on Gaussian Processes (AT-GP), which can be used to adapt the transfer learning schemes by automatically estimating the similarity between a source and a target task. The main contribution of our work is that we propose a new semi-parametric transfer kernel for transfer learning from a Bayesian perspective, and propose to learn the model with respect to the target task, rather than all tasks as in multi-task learning. We can formulate the transfer learning problem as a unified Gaussian Process (GP) model. The adaptive transfer ability of our approach is verified on both synthetic and real-world datasets.

Sentiment classification aims to automatically predict sentiment polarity (e.g., positive or negative) of users publishing sentiment data (e.g., reviews, blogs). Although traditional classification algorithms can be used to train sentiment classifiers from manually labeled text data, the labeling work can be time-consuming and expensive. Meanwhile, users often use some different words when they express sentiment in different domains. If we directly apply a classifier trained in one domain to other domains, the performance will be very low due to the differences between these domains. In this work, we develop a general solution to sentiment classification when we do not have any labels in a target domain but have some labeled data in a different domain, regarded as source domain. In this cross-domain sentiment classification setting, to bridge the gap between the domains, we propose a spectral feature alignment (SFA) algorithm to align domain-specific words from different domains into unified clusters, with the help of domain-independent words as a bridge. In this way, the clusters can be used to reduce the gap between domain-specific words of the two domains, which can be used to train sentiment classifiers in the target domain accurately. Compared to previous approaches, SFA can discover a robust representation for cross-domain data by fully exploiting the relationship between the domain-specific and domain-independent words via simultaneously co-clustering them in a common latent space. We perform extensive experiments on two real world datasets, and demonstrate that SFA significantly outperforms previous approaches to cross-domain sentiment classification.

A major assumption in many machine learning and data mining algorithms is that the training and future data must be in the same feature space and have the same distribution. However, in many real-world applications, this assumption may not hold. For example, we sometimes have a classification task in one domain of interest, but we only have sufficient training data in another domain of interest, where the latter data may be in a different feature space or follow a different data distribution. In such cases, knowledge transfer, if done successfully, would greatly improve the performance of learning by avoiding much expensive data-labeling efforts. In recent years, transfer learning has emerged as a new learning framework to address this problem. This survey focuses on categorizing and reviewing the current progress on transfer learning for classification, regression, and clustering problems. In this survey, we discuss the relationship between transfer learning and other related machine learning techniques such as domain adaptation, multitask learning and sample selection bias, as well as covariate shift. We also explore some potential future issues in transfer learning research.

Recognizing and understanding the activities of people from sensor readings is an important task in ubiquitous computing. Activity recognition is also a particularly difficult task because of the inherent uncertainty and complexity of the data collected by the sensors. Many researchers have tackled this problem in an overly simplistic setting by assuming that users often carry out single activities one at a time or multiple activities consecutively, one after another. However, so far there has been no formal exploration on the degree in which humans perform concurrent or interleaving activities, and no thorough study on how to detect multiple goals in a real world scenario. In this article, we ask the fundamental questions of whether users often carry out multiple concurrent and interleaving activities or single activities in their daily life, and if so, whether such complex behavior can be detected accurately using sensors. We define several classes of complexity levels under a goal taxonomy that describe different granularities of activities, and relate the recognition accuracy with different complexity levels or granularities. We present a theoretical framework for recognizing multiple concurrent and interleaving activities, and evaluate the framework in several real-world ubiquitous computing environments.

With the proliferation of wireless and sensor techniques, data can be shared conveniently through the air. However, wireless communication is vulnerable since unauthorized machine may try to intrude a server without being physically connected. In this paper, we wish to control the communication between a wireless client and the infrastructure based on the client location. Our idea is to implement a digital wall, which is a user-defined boundary so that access is allowed within the boundary and denied outside the boundary. To do this, we need to do accurate location estimation since the decision around the boundary line is critical. Furthermore, computational efficiency is also important since we need to reduce computation cost so as to save power energy. In this paper, we propose k-nearest-neighbor (KNN) based method to determine the location of a mobile client based on received signal strength (RSS) values. We further use information gain as a feature selection criterion to reduce the estimation time. We study how the performance may be affected when the controlled area is confined by physical or virtual walls. Experimental results show that we can well distinguish whether a client device is located within an expected digital wall.

Accurate localization of mobile objects is a major research problem in sensor networks and an important data mining application. Specifically, the localization problem is to determine the location of a client device accurately given the radio signal strength values received at the client device from multiple beacon sensors or access points. Conventional data mining and machine learning methods can be applied to solve this problem. However, all of them require large amounts of labeled training data, which can be quite expensive. In this paper, we propose a probabilistic semi supervised learning approach to reduce the calibration effort and increase the tracking accuracy. Our method is based on semi-supervised conditional random fields which can enhance the learned model from a small set of training data with abundant unlabeled data effectively. To make our method more efficient, we exploit a Generalized EM algorithm coupled with domain constraints. We validate our method through extensive experiments in a real sensor network using Crossbow MICA2 sensors. The results demonstrate the advantages of methods compared to other state-of-the-art object-tracking algorithms.