One global model in federated learning (FL) might not be sufficient to serve many clients with non-IID tasks and distributions. Despite recent advances in FL to train multiple global models for better personalization, they only provide limited model choices to clients, so local finetuning of multiple models is still indispensable. This paper proposes a novel ``FedMerge'' approach that can create a single personalized model per client by simply merging multiple global models with automatically optimized and customized weights. We formulate this problem as a joint optimization of global models and the merging weights per client. Unlike existing FL approaches, where the server broadcasts one or multiple global models to all clients, the server only needs to send a customized, merged model to each client. Moreover, instead of periodically interrupting the local training and re-initializing it to a global model, the merged model aligns better with each client's task and data distribution, smoothening the local-global gap between consecutive rounds caused by client drift. We evaluate FedMerge on different non-IID settings applied to various domains with diverse tasks and data types, in which FedMerge consistently outperforms existing FL approaches, including clustering-based and mixture-of-experts (MoE) based methods.

Multi-agent systems (MAS) based on large language models (LLMs) have emerged as a powerful solution for dealing with complex problems across diverse domains. The effectiveness of MAS is critically dependent on its collaboration topology, which has become a focal point for automated design research. However, existing approaches are fundamentally constrained by their reliance on a template graph modification paradigm with a predefined set of agents and hard-coded interaction structures, significantly limiting their adaptability to task-specific requirements. To address these limitations, we reframe MAS design as a conditional autoregressive graph generation task, where both the system composition and structure are designed jointly. We propose ARG-Designer, a novel autoregressive model that operationalizes this paradigm by constructing the collaboration graph from scratch. Conditioned on a natural language task query, ARG-Designer sequentially and dynamically determines the required number of agents, selects their appropriate roles from an extensible pool, and establishes the optimal communication links between them. This generative approach creates a customized topology in a flexible and extensible manner, precisely tailored to the unique demands of different tasks. Extensive experiments across six diverse benchmarks demonstrate that ARG-Designer not only achieves state-of-the-art performance but also enjoys significantly greater token efficiency and enhanced extensibility.

Fairness-aware federated graph neural networks (FedGNNs) necessitate consideration of both the server and the clients. However, fairness-aware methods struggle to enhance dual-perspective (i.e., server and clients) fairness without sacrificing utility due to the distributed learning framework. As a consequence, the utility sacrifices of fairness-aware graph learning methods are even exacerbated in federated frameworks. In this work we propose F3GL, a dual-perspective fairness federated graph learning method that enhances both global (for the server) and local fairness (for clients) while preserving utility. Through theoretical analysis, we delineate the similarity between original sensitive features and those after convolution under different spectra. Our findings reveal that only the principal eigenvalue contributes to enhancing this similarity. Moreover, our theoretical analysis applies universally to both clients and servers. Specifically, employing a specialized eigenvalue selection strategy allows for effective optimization of both local and global fairness. Drawing on these insights, we improve dual-perspective fairness through the lens of spectral graph theory without sacrificing utility. Experimental results on two real-world datasets show the superiority of F3GL over existing baselines.

Knowledge graphs (KGs) store enormous facts as relationships between entities. Due to the long-tailed distribution of relations and the incompleteness of KGs, there is growing interest in few-shot knowledge graph completion (FKGC). Existing FKGC methods often assume the existence of all entities in KGs, which may not be practical since new relations and entities can emerge over time. Therefore, we focus on a more challenging task called inductive few-shot knowledge graph completion (I-FKGC), where both relations and entities during the test phase are unknown before. Inspired by the idea of inductive reasoning, we cast I-FKGC as an inductive reasoning problem. Specifically, we propose a novel Graph Stochastic Neural Process ( GS-NP ) approach, which consists of two major modules. In the first module, to obtain a generalized hypothesis (e.g., shared subgraph), we present a neural process-based hypothesis extractor that models the joint distribution of hypothesis, from which we can sample a hypothesis for predictions. In the second module, based on the hypothesis, we propose a graph stochastic attention-based predictor to test if the triple in the query set aligns with the extracted hypothesis. Meanwhile, the predictor can generate an explanatory subgraph identified by the hypothesis. Finally, the training of these two modules is seamlessly combined into a unified objective function, of which the effectiveness is verified by theoretical analyses as well as empirical studies. Extensive experiments on three public datasets demonstrate that our method outperforms existing methods and derives new state-of-the-art performance.

Spatio-temporal video grounding (STVG) aims to precisely locate a spatio-temporal tube in an untrimmed video corresponding to a given language description. Many existing methods decouple spatial and temporal grounding as separate tasks, missing the strong interdependencies between the two, which are crucial for accurately aligning spatial regions (such as objects) with their motion over time. Thus, to enhance spatio-temporal associations, we introduce a new Prior-Driven Transformer Network (PDTNet) with predicted temporal boundaries as priors to guide object bounding boxes for improved spatial grounding over time. Firstly, PDTNet employs a temporal prior, termed reference query, to enhance discriminability between language-related and language-irrelevant visual content, improving temporal boundary localization. Further, the context within predicted temporal boundaries serves as another prior knowledge to modulate spatial features. We also introduce a prediction-aware Gaussian prior to precise object localization. This ensures consistent tube construction and accurate object localization. Extensive experiments on STVG benchmarks validate the effectiveness of PDTNet. Code is available at https://github.com/tongzhang111/PDTNet .

Training a general-purpose time series foundation models with robust generalization capabilities across diverse applications from scratch is still an open challenge. Efforts are primarily focused on fusing cross-domain time series datasets to extract shared subsequences as tokens for training models on Transformer architecture. However, due to significant statistical heterogeneity across domains, this cross-domain fusing approach doesn't work effectively as the same as fusing texts and images. To tackle this challenge, this paper proposes a novel federated learning approach to address the heterogeneity in time series foundation models training, namely FFTS. Specifically, each data-holding organization is treated as an independent client in a collaborative learning framework with federated settings, and then many client-specific local models will be trained to preserve the unique characteristics per dataset. Moreover, a new regularization mechanism will be applied to both client-side and server-side, thus to align the shared knowledge across heterogeneous datasets from different domains. Extensive experiments on benchmark datasets demonstrate the effectiveness of the proposed federated learning approach. The newly learned time series foundation models achieve superior generalization capabilities on cross-domain time series analysis tasks, including forecasting, imputation, and anomaly detection.

Federated Collaborative Filtering (FedCF) is an emerging field focused on developing a new recommendation framework with preserving privacy in a federated setting. Existing FedCF methods typically combine distributed Collaborative Filtering (CF) algorithms with privacy-preserving mechanisms, and then preserve personalized information into a user embedding vector. However, the user embedding is usually insufficient to preserve the rich information of the fine-grained personalization across heterogeneous clients. This paper proposes a novel personalized FedCF method by preserving users' personalized information into a latent variable and a neural model simultaneously. Specifically, we decompose the modeling of user knowledge into two encoders, each designed to capture shared knowledge and personalized knowledge separately. A personalized gating network is then applied to balance personalization and generalization between the global and local encoders. Moreover, to effectively train the proposed framework, we model the CF problem as a specialized Variational AutoEncoder (VAE) task by integrating user interaction vector reconstruction with missing value prediction. The decoder is trained to reconstruct the implicit feedback from items the user has interacted with, while also predicting items the user might be interested in but has not yet interacted with. Experimental results on benchmark datasets demonstrate that the proposed method outperforms other baseline methods, showcasing superior performance.

State-of-the-art brain graph analysis methods fail to fully encode the small-world architecture of brain graphs (accompanied by the presence of hubs and functional modules), and therefore lack biological plausibility to some extent. This limitation hinders their ability to accurately represent the brain's structural and functional properties, thereby restricting the effectiveness of machine learning models in tasks such as brain disorder detection. In this work, we propose a novel Biologically Plausible Brain Graph Transformer (BioBGT) that encodes the small-world architecture inherent in brain graphs. Specifically, we present a network entanglement-based node importance encoding technique that captures the structural importance of nodes in global information propagation during brain graph communication, highlighting the biological properties of the brain structure. Furthermore, we introduce a functional module-aware self-attention to preserve the functional segregation and integration characteristics of brain graphs in the learned representations. Experimental results on three benchmark datasets demonstrate that BioBGT outperforms state-of-the-art models, enhancing biologically plausible brain graph representations for various brain graph analytical tasks

Time series analysis is crucial across various fields like energy, environment, transportation, finance and health. Deep learning has significantly advanced this field, particularly, the Time Series Foundation Model (TSFM) excels in multiple domains due to extensive pre-training. In this work, we focus on TSFM’s challenges in medical practice: limited computing resources and medical data privacy. TSFM variants include fine-tuned models and those pre-trained for rapid deployment on diverse data. There may not be enough computing resources to train physiological signals locally in hospitals, and generalized TSFM is still inferior to task-specific methods on private, imbalanced local data. To address this, we propose PhysioPFM, a framework for efficiently personalizing TSFM. Our approach involves low-rank pre-training on public datasets, generator training by trained LoRA weights, and efficient weight generation via local data. Experimental results demonstrate that integrating generated models with TSFM enhances performance, and transferability, and reduces the need for additional sensitive data training.

Loss reweighting has shown significant benefits for machine unlearning with large language models (LLMs). However, their exact functionalities are left unclear and the optimal strategy remains an open question, thus impeding the understanding and improvement of existing methodologies. In this paper, we identify two distinct goals of loss reweighting, namely, Saturation and Importance---the former indicates that those insufficiently optimized data should be emphasized, while the latter stresses some critical data that are most influential for loss minimization. To study their usefulness, we design specific reweighting strategies for each goal and evaluate their respective effects on unlearning. We conduct extensive empirical analyses on well-established benchmarks, and summarize some important observations as follows: (i) Saturation enhances efficacy more than importance-based reweighting, and their combination can yield additional improvements. (ii) Saturation typically allocates lower weights to data with lower likelihoods, whereas importance-based reweighting does the opposite. (iii) The efficacy of unlearning is also largely influenced by the smoothness and granularity of the weight distributions. Based on these findings, we propose SatImp, a simple reweighting method that combines the advantages of both saturation and importance. Empirical results on extensive datasets validate the efficacy of our method, potentially bridging existing research gaps and indicating directions for future research. Our code is available at https://github.com/tmlr-group/SatImp.

Effectively leveraging private datasets remains a significant challenge in developing foundation models. Federated Learning (FL) has recently emerged as a collaborative framework that enables multiple users to fine-tune these models while mitigating data privacy risks. Meanwhile, Low-Rank Adaptation (LoRA) offers a resource-efficient alternative for fine-tuning foundation models by dramatically reducing the number of trainable parameters. This survey examines how LoRA has been integrated into federated fine-tuning for foundation models—an area we term FedLoRA—by focusing on three key challenges: distributed learning, heterogeneity, and efficiency. We further categorize existing work based on the specific methods used to address each challenge. Finally, we discuss open research questions and highlight promising directions for future investigation, outlining the next steps for advancing FedLoRA.

Representation learning on electronic health records (EHRs) plays a vital role in downstream medical prediction tasks. Although natural language processing techniques, such as recurrent neural networks, and self-attention, have been adapted for learning medical representations from hierarchical, time-stamped EHR data, they often struggle when either general or task-specific data are limited. Recent efforts have attempted to mitigate this challenge by incorporating medical ontologies (i.e., knowledge graphs) into self-supervised tasks like diagnosis prediction. However, two main issues remain: (1) small and uniform ontologies that lack diversity for robust learning, and (2) insufficient attention to the critical contexts or dependencies underlying patient journeys, which could further enhance ontology-based learning. To address these gaps, we propose MIPO (Mutual Integration of Patient Journey and Medical Ontology), a robust end-to-end framework that employs a Transformer-based architecture for representation learning. MIPO emphasizes task-specific representation learning through a sequential diagnosis prediction task, while also incorporating an ontology-based disease-typing task. A graph-embedding module is introduced to integrate information from patient visit records, thus alleviating data insufficiency. This setup creates a mutually reinforcing loop, where both patient-journey embedding and ontology embedding benefit from each other. We validate MIPO on two real-world benchmark datasets, showing that it consistently outperforms baseline methods under both sufficient and limited data conditions. Furthermore, the resulting diagnosis embeddings offer improved interpretability, underscoring the promise of MIPO for real-world healthcare applications.

Can we build accurate world models out of large language models (LLMs)? How can world models benefit LLM agents? The gap between the prior knowledge of LLMs and the specified environment's dynamics usually bottlenecks LLMs' performance as world models. To bridge the gap, we propose a training-free "world alignment" that learns an environment's symbolic knowledge complementary to LLMs. The symbolic knowledge covers action rules, knowledge graphs, and scene graphs, which are extracted by LLMs from exploration trajectories and encoded into executable codes to regulate LLM agents' policies. We further propose an RL-free, model-based agent "WALL-E" through the model-predictive control (MPC) framework. Unlike classical MPC requiring costly optimization on the fly, we adopt an LLM agent as an efficient look-ahead optimizer of future steps' actions by interacting with the neurosymbolic world model. While the LLM agent's strong heuristics make it an efficient planner in MPC, the quality of its planned actions is also secured by the accurate predictions of the aligned world model. They together considerably improve learning efficiency in a new environment. On open-world challenges in Mars (Minecraft like) and ALFWorld (embodied indoor environments), WALL-E significantly outperforms existing methods, e.g., surpassing baselines in Mars by 16.1%–51.6% of success rate and by at least 61.7% in score. In ALFWorld, it achieves a new record 98% success rate after only 4 iterations.

Abstract Recommender systems have been widely deployed in various real-world applications to help users identify content of interest from massive amounts of information. Traditional recommender systems work by collecting user-item interaction data in a cloud-based data center and training a centralized model to perform the recommendation service. However, such cloud-based recommender systems (CloudRSs) inevitably suffer from excessive resource consumption, response latency, as well as privacy and security risks concerning both data and models. Recently, driven by the advances in storage, communication, and computation capabilities of edge devices, there has been a shift of focus from CloudRSs to on-device recommender systems (DeviceRSs), which leverage the capabilities of edge devices to minimize centralized data storage requirements, reduce the response latency caused by communication overheads, and enhance user privacy and security by localizing data processing and model training. Despite the rapid rise of DeviceRSs, there is a clear absence of timely literature reviews that systematically introduce, categorize and contrast these methods. To bridge this gap, we aim to provide a comprehensive survey of DeviceRSs, covering three main aspects: (1) the deployment and inference of DeviceRSs, exploring how large recommendation models can be compressed and utilized within resource-constrained on-device environments; (2) the training and update of DeviceRSs, discussing how local data can be leveraged for model optimization on the device side; (3) the security and privacy of DeviceRSs, unveiling their potential vulnerability to malicious attacks and defensive strategies to safeguard these systems. Furthermore, we provide a fine-grained and systematic taxonomy of the methods involved in each aspect, followed by a discussion regarding challenges and future research directions. This is the first comprehensive survey on DeviceRSs that covers a spectrum of tasks to fit various needs. We believe this survey will help readers understand the current research status in this field, equip them with relevant technical foundations, and stimulate new research ideas for developing DeviceRSs.

Repetitive Negative Sequential Patterns (RNSPs) can provide critical insights into the importance of sequences. However, most current RNSP mining methods require users to set an appropriate support threshold to obtain the expected number of patterns, which is a very difficult task for the users without prior experience . To address this issue, we propose a new algorithm, TK-RNSP, to mine the Top- K RNSPs with the highest support, without the need to set a support threshold. In detail, we achieve a significant breakthrough by proposing a series of definitions that enable RNSP mining to satisfy anti-monotonicity. Then, we propose a bitmap-based Depth-First Backtracking Search (DFBS) strategy to decrease the heavy computational burden by increasing the speed of support calculation. Finally, we propose the algorithm TK-RNSP in an one-stage process, which can effectively reduce the generation of unnecessary patterns and improve computational efficiency comparing to those two-stage process algorithms. To the best of our knowledge, TK-RNSP is the first algorithm to mine Top- K RNSPs. Extensive experiments on eight datasets show that TK-RNSP has better flexibility and efficiency to mine Top- K RNSPs.

Traffic forecasting plays a crucial role in establishing an Intelligent Transportation System (ITS) by providing essential insights. Existing traffic forecasting relies on the assumption that there is a hidden invariant spatial-temporal pattern in the large-scale dataset. However, the traffic patterns are easily influenced by many unpredictable external factors, such as policy interventions and climate changes. Due to the dynamic nature of these exogenous factors, the traffic network's spatial-temporal patterns are also changed, thus impacting the performance of traffic forecasting models. Thus, there is an urgent need to rethink the traffic forecasting model in a fast-adaptive manner. To solve this challenge, this paper proposes an Adaptive Spatio-Temporal Context Learning framework named ASTCL, which achieves desired forecasting accuracy using daily basis traffic data collected from dozens of sensors. ASTCL constructs adaptive spatio-temporal contexts for target locations in the traffic network and generates dynamic sequence graphs based on semantic similarities. The adaptive contexts aggregate valuable information from available data, while the graphs reveal dynamic trends in traffic properties. Further, ASTCL introduces a joint convolution and attention mechanism to model intricate spatio-temporal relationships from multiple perspectives. Extensive experiments conducted on four real-world datasets demonstrate that ASTCL achieves remarkable fast adaptability and outperforms other state-of-the-art methods by a significant margin.

Fine-grained urban flow inference (FUFI) is crucial for traffic management, as it infers high-resolution urban flow maps from coarse-grained observations. Existing FUFI methods typically focus on a single city and rely on comprehensive training with large-scale datasets to achieve precise inferences. However, data availability in developing cities may be limited, posing challenges to the development of well-performing models. To address this issue, we propose cross-city fine-grained urban flow inference, which aims to transfer spatio-temporal knowledge from data-rich cities to data-scarce areas using meta-transfer learning. This paper devises a Spatio-Temporal Deviation Alignment (STDA) framework to mitigate spatio-temporal distribution deviations and urban structural deviations between multiple source cities and the target city. Furthermore, STDA presents a cross-city normalization method that adaptively combines batch and instance normalization to maintain consistency between city-variant and city-invariant features. Besides, we design an urban structure alignment module to align spatial topological differences across cities. STDA effectively reduces distribution and structural deviations among different datasets while avoiding negative transfer. Extensive experiments conducted on three real-world datasets demonstrate that STDA consistently outperforms state-of-the-art baselines.

Dynamic material handling (DMH) involves the assignment of dynamically arriving material transporting tasks to suitable vehicles in real time for minimizing makespan and tardiness. In real-world scenarios, historical task records are usually available, which enables the training of a decision policy on multiple instances consisting of historical records. Recently, reinforcement learning (RL) has been applied to solve DMH. Due to the occurrence of dynamic events such as new tasks, adaptability is highly required. Solving DMH is challenging since constraints, including task delay, should be satisfied. A feedback is received only when all tasks are served, which leads to sparse reward. Besides, making the best use of limited computational resources and historical records for training a robust policy is crucial. The time allocated to different problem instances would highly impact the learning process. To tackle those challenges, this article proposes a novel adaptive constrained evolutionary RL (ACERL) approach, which maintains a population of actors for diverse exploration. ACERL accesses each actor for tackling sparse rewards and constraint violation to restrict the behavior of the policy. Moreover, ACERL adaptively selects the most beneficial training instances for improving the policy. Extensive experiments on eight training and eight unseen test instances demonstrate the outstanding performance of ACERL compared with several state-of-the-art algorithms. Policies trained by ACERL can schedule the vehicles while fully satisfying the constraints. Additional experiments on 40 unseen noised instances show the robust performance of ACERL. Cross validation further presents the overall effectiveness of ACREL. Besides, a rigorous ablation study highlights the coordination and benefits of each ingredient of ACERL.

Localizing text in low-light environments is challenging due to visual degradations. Although a straightforward solution involves a two-stage pipeline with low-light image enhancement (LLE) as the initial step followed by detection, LLE is primarily designed for human vision rather than machine vision and can accumulate errors. In this work, we propose an efficient and effective single-stage approach for localizing text in the dark that circumvents the need for LLE. We introduce a constrained learning module as an auxiliary mechanism during the training stage of the text detector. This module is designed to guide the text detector in preserving textual spatial features amidst feature map resizing, thus minimizing the loss of spatial information in texts under low-light visual degradations. Specifically, we incorporate spatial reconstruction and spatial semantic constraints within this module to ensure the text detector acquires essential positional and contextual range knowledge. Our approach enhances the original text detector's ability to identify text's local topological features using a dynamic snake feature pyramid network and adopts a bottom-up contour shaping strategy with a novel rectangular accumulation technique for accurate delineation of streamlined text features. In addition, we present a comprehensive low-light dataset for arbitrary-shaped text, encompassing diverse scenes and languages. Notably, our method achieves state-of-the-art results on this low-light dataset and exhibits comparable performance on standard normal light datasets. The code and dataset will be released.

Graph anomaly detection (GAD), which aims to identify abnormal nodes that differ from the majority within a graph, has garnered significant attention. However, current GAD methods necessitate training specific to each dataset, resulting in high training costs, substantial data requirements, and limited generalizability when being applied to new datasets and domains. To address these limitations, this paper proposes ARC, a generalist GAD approach that enables a ``one-for-all'' GAD model to detect anomalies across various graph datasets on-the-fly. Equipped with in-context learning, ARC can directly extract dataset-specific patterns from the target dataset using few-shot normal samples at the inference stage, without the need for retraining or fine-tuning on the target dataset. ARC comprises three components that are well-crafted for capturing universal graph anomaly patterns: 1) smoothness-based feature **A**lignment module that unifies the features of different datasets into a common and anomaly-sensitive space; 2) ego-neighbor **R**esidual graph encoder that learns abnormality-related node embeddings; and 3) cross-attentive in-**C**ontext anomaly scoring module that predicts node abnormality by leveraging few-shot normal samples. Extensive experiments on multiple benchmark datasets from various domains demonstrate the superior anomaly detection performance, efficiency, and generalizability of ARC.

This paper demonstrates that pre-trained language models (PLMs) are strong foundation models for on-device meteorological variable modeling. We present LM-Weather, a generic approach to taming PLMs, that have learned massive sequential knowledge from the universe of natural language databases, to acquire an immediate capability to obtain highly customized models for heterogeneous meteorological data on devices while keeping high efficiency. Concretely, we introduce a lightweight personalized adapter into PLMs and endows it with weather pattern awareness. During communication between clients and the server, low-rank-based transmission is performed to effectively fuse the global knowledge among devices while maintaining high communication efficiency and ensuring privacy. Experiments on real-wold dataset show that LM-Weather outperforms the state-of-the-art results by a large margin across various tasks (e.g., forecasting and imputation at different scales). We provide extensive and in-depth analyses experiments, which verify that LM-Weather can (1) indeed leverage sequential knowledge from natural language to accurately handle meteorological sequence, (2) allows each devices obtain highly customized models under significant heterogeneity, and (3) generalize under data-limited and out-of-distribution (OOD) scenarios.

In _cross-domain few-shot classification_ (CFC), recent works mainly focus on adapting a simple transformation head on top of a frozen pre-trained backbone with few labeled data to project embeddings into a task-specific metric space where classification can be performed by measuring similarities between image instance and prototype representations. Technically, an _assumption_ implicitly adopted in such a framework is that the prototype and image instance embeddings share the same representation transformation. However, in this paper, we find that there naturally exists a gap, which resembles the modality gap, between the prototype and image instance embeddings extracted from the frozen pre-trained backbone, and simply applying the same transformation during the adaptation phase constrains exploring the optimal representation distributions and shrinks the gap between prototype and image representations. To solve this problem, we propose a simple yet effective method, _contrastive prototype-image adaptation_ (CoPA), to adapt different transformations for prototypes and images similarly to CLIP by treating prototypes as text prompts. Extensive experiments on Meta-Dataset demonstrate that CoPA achieves the _state-of-the-art_ performance more efficiently. Meanwhile, further analyses also indicate that CoPA can learn better representation clusters, enlarge the gap, and achieve the minimum validation loss at the enlarged gap.

Despite the recent success of deep neural networks in spatial-temporal forecasting, existing methods suffer from distribution shifts between the training and test data, failing to address the non-stationary and abrupt changes at test time. To solve this problem, we propose a novel test-time training framework for spatial-temporal forecasting. Instead of employing a fixed trained model, we adapt the trained model with only one or a mini-batch of test examples to address the test data shifts. The unique spatial structure with hundreds of geographical locations offers an effective batch size to explore the test-time distribution and avoid overfitting.

Multi-graph Multi-label learning (Mgml) aims to classify a set of objects of interest, such as text or images, using a bag-of-graphs representation. PreviousMgmlworks have limitations as they only learn labels at the bag level, lose structural information in learning by transferring graphs into instances, and cannot handle noisy labels. This paper presents a robustcoarseandfine-grainedNoise Multi-graph Multi-label (cfMGNML) learning framework that builds the learning model over the graphs and empowers label prediction at both thecoarse(bag) andfine-grained(graph in each bag) levels with noisy labels. To identify label noise, a label probability matrix is defined to act on the scoring function of each label, with a higher probability value indicating that the label is more likely to be the corresponding graph or bag label. The problem is regularized with the manifold constraint on the label probability matrix to preserve local relationships within the data and uncover its essential manifold structure. Meanwhile, a thresholding rank-loss objective is proposed to rank the labels for the graphs and bags and minimize the hamming loss at one step simultaneously. To tackle the non-convex optimization problem, an effective sub- gradient descent algorithm is developed. Experiments over various datasets demonstrate the proposed method achieves superior performance than the state-of-the-art algorithms.

In this study, we present a transductive inference approach on that reward information propagation graph, which enables the effective estimation of rewards for unlabelled data in offline reinforcement learning. Reward inference is the key to learning effective policies in practical scenarios, while direct environmental interactions are either too costly or unethical and the reward functions are rarely accessible, such as in healthcare and robotics. Our research focuses on developing a reward inference method based on the contextual properties of information propagation on graphs that capitalizes on a constrained number of human reward annotations to infer rewards for unlabelled data. We leverage both the available data and limited reward annotations to construct a reward propagation graph, wherein the edge weights incorporate various influential factors pertaining to the rewards. Subsequently, we employ the constructed graph for transductive reward inference, thereby estimating rewards for unlabelled data. Furthermore, we establish the existence of a fixed point during several iterations of the transductive inference process and demonstrate its at least convergence to a local optimum. Empirical evaluations on locomotion and robotic manipulation tasks validate the effectiveness of our approach. The application of our inferred rewards improves the performance in offline reinforcement learning tasks.

Graph neural networks (GNNs) have shown their superiority in modeling graph data. Owing to the advantages of federated learning, federated graph learning (FGL) enables clients to train strong GNN models in a distributed manner without sharing their private data. A core challenge in federated systems is the non-IID problem, which also widely exists in real-world graph data. For example, local data of clients may come from diverse datasets or even domains, e.g., social networks and molecules, increasing the difficulty for FGL methods to capture commonly shared knowledge and learn a generalized encoder. From real-world graph datasets, we observe that some structural properties are shared by various domains, presenting great potential for sharing structural knowledge in FGL. Inspired by this, we propose FedStar, an FGL framework that extracts and shares the common underlying structure information for inter-graph federated learning tasks. To explicitly extract the structure information rather than encoding them along with the node features, we define structure embeddings and encode them with an independent structure encoder. Then, the structure encoder is shared across clients while the feature-based knowledge is learned in a personalized way, making FedStar capable of capturing more structure-based domain-invariant information and avoiding feature misalignment issues. We perform extensive experiments over both cross-dataset and cross-domain non-IID FGL settings, demonstrating the superiority of FedStar.

Xiuying Chen, Guodong Long, Chongyang Tao, Mingzhe Li, Xin Gao, Chengqi Zhang, Xiangliang Zhang. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

As a critical mission of intelligent transportation systems, urban flow prediction (UFP) benefits in many city services including trip planning, congestion control, and public safety. Despite the achievements of previous studies, limited efforts have been observed on simultaneous investigation of the heterogeneity in both space and time aspects. That is, regional correlations would be variable at different timestamps. In this paper, we propose a spatio-temporal learning framework with mask and contrast enhancements to capture spatio-temporal variabilities among city regions. We devise a mask-enhanced pre-training task to learn latent correlations across the spatial and temporal dimensions, and then a graph-based method is developed to extract the significance of regions by using the inter-regional attention weights. To further acquire contrastive correlations of regions, we elaborate a pre-trained contrastive learning task with the global-local cross-attention mechanism. Thereafter, two well-trained encoders have strong capability to capture latent spatio-temporal representations for the flow forecasting with time-varying. Extensive experiments conducted on real-world urban flow datasets demonstrate that our method compares favorably with other state-of-the-art models.

Data similarity computation is a fundamental research topic which underpins many high-level applications based on similarity measures. However, the exact similarity computation has become daunting in large-scale real-world scenarios. Currently, MinHash is a popular technique for efficiently estimating the Jaccard similarity of binary sets and, furthermore, weighted MinHash is utilized to estimate the generalized Jaccard similarity of weighted sets. This review focuses on categorizing and discussing the existing works of weighted MinHash algorithms. Also, we have developed a Python toolbox for the algorithms, and released it in our github.

Distribution shift (e.g., task or domain shift) in continual learning (CL) usually results in catastrophic forgetting of previously learned knowledge. Although it can be alleviated by repeatedly replaying buffered data, the every-step replay is time-consuming. In this paper, we study which modules in neural networks are more prone to forgetting by investigating their training dynamics during CL. Our proposed metrics show that only a few modules are more task-specific and sensitive to task change, while others can be shared across tasks as common knowledge. Hence, we attribute forgetting mainly to the former and find that finetuning them only on a small buffer at the end of any CL method can bring non-trivial improvement. Due to the small number of finetuned parameters, such ''Forgetting Prioritized Finetuning (FPF)'' is efficient in computation. We further propose a more efficient and simpler method that entirely removes the every-step replay and replaces them by only $k$-times of FPF periodically triggered during CL. Surprisingly, this ''$k$-FPF'' performs comparably to FPF and outperforms the SOTA CL methods but significantly reduces their computational overhead and cost. In experiments on several benchmarks of class- and domain-incremental CL, FPF consistently improves existing CL methods by a large margin, and $k$-FPF further excels in efficiency without degrading the accuracy. We also empirically studied the impact of buffer size, epochs per task, and finetuning modules on the cost and accuracy of our methods.

Federated recommendation is a new Internet service architecture that aims to provide privacy-preserving recommendation services in federated settings. Existing solutions are used to combine distributed recommendation algorithms and privacy-preserving mechanisms. Thus it inherently takes the form of heavyweight models at the server and hinders the deployment of on-device intelligent models to end-users. This paper proposes a novel Personalized Federated Recommendation (PFedRec) framework to learn many user-specific lightweight models to be deployed on smart devices rather than a heavyweight model on a server. Moreover, we propose a new dual personalization mechanism to effectively learn fine-grained personalization on both users and items. The overall learning process is formulated into a unified federated optimization framework. Specifically, unlike previous methods that share exactly the same item embeddings across users in a federated system, dual personalization allows mild finetuning of item embeddings for each user to generate user-specific views for item representations which can be integrated into existing federated recommendation methods to gain improvements immediately. Experiments on multiple benchmark datasets have demonstrated the effectiveness of PFedRec and the dual personalization mechanism. Moreover, we provide visualizations and in-depth analysis of the personalization techniques in item embedding, which shed novel insights on the design of recommender systems in federated settings. The code is available.

Heterogeneous federated learning without assuming any structure is challenging due to the conflicts among non-identical data distributions of clients. In practice, clients often comprise near-homogeneous clusters so training a server-side model per cluster mitigates the conflicts. However, FL with client clustering often suffers from “clustering collapse'', i.e., one cluster's model excels on increasing clients, and reduces to single-model FL. Moreover, cluster-wise models hinder knowledge sharing between clusters and each model depends on fewer clients. Furthermore, the static clustering assumption on data may not hold for dynamically changing models, which are sensitive to cluster imbalance/initialization or outliers. To address these challenges, we propose ''Clustered Additive Modeling (CAM)'', which applies a globally shared model $\Theta_g$ on top of the cluster-wise models $\Theta_{1:K}$, i.e., $y=h(x;\Theta_g)+f(x;\Theta_k)$ for clients of cluster-$k$. The global model captures the features shared by all clusters so $\Theta_{1:K}$ are enforced to focus on the difference among clusters. To train CAM, we develop a novel Fed-CAM algorithm that alternates between client clustering and training global/cluster models to predict the residual of each other. We can easily modify any existing clustered FL methods by CAM and significantly improve their performance without ‘’clustering collapse'' in different non-IID settings. We also provide a convergence analysis of Fed-CAM algorithm.

Graph convolutional networks are becoming indispensable for deep learning from graph-structured data. Most of the existing graph convolutional networks share two big shortcomings. First, they are essentially low-pass filters, thus the potentially useful middle and high frequency band of graph signals are ignored. Second, the bandwidth of existing graph convolutional filters is fixed. Parameters of a graph convolutional filter only transform the graph inputs without changing the curvature of a graph convolutional filter function. In reality, we are uncertain about whether we should retain or cut off the frequency at a certain point unless we have expert domain knowledge. In this paper, we propose Automatic Graph Convolutional Networks (AutoGCN) to capture the full spectrum of graph signals and automatically update the bandwidth of graph convolutional filters. While it is based on graph spectral theory, our AutoGCN is also localized in space and has a spatial form. Experimental results show that AutoGCN achieves significant improvement over baseline methods which only work as low-pass filters.

Urban traffic forecasting is the cornerstone of the intelligent transportation system (ITS). Existing methods focus on spatial-temporal dependency modeling, while two intrinsic properties of the traffic forecasting problem are overlooked. First, the complexity of diverse forecasting tasks is nonuniformly distributed across various spaces (e.g., suburb versus downtown) and times (e.g., rush hour versus off-peak). Second, the recollection of past traffic conditions is beneficial to the prediction of future traffic conditions. Based on these properties, we propose a bidirectional spatial-temporal adaptive transformer (Bi-STAT) for accurate traffic forecasting. Bi-STAT adopts an encoder-decoder architecture, where both the encoder and the decoder maintain a spatial-adaptive transformer and a temporal-adaptive transformer structure. Inspired by the first property, each transformer is designed to dynamically process the traffic streams according to their task complexities. Specifically, we realize this by the recurrent mechanism with a novel dynamic halting module (DHM). Each transformer performs iterative computation with shared parameters until DHM emits a stopping signal. Motivated by the second property, Bi-STAT utilizes one decoder to perform the present → past recollection task and the other decoder to perform the present → future prediction task. The recollection task supplies complementary information to assist and regularize the prediction task for a better generalization. Through extensive experiments, we show the effectiveness of each module in Bi-STAT and demonstrate the superiority of Bi-STAT over the state-of-the-art baselines on four benchmark datasets. The code is available at https://github.com/chenchl19941118/Bi-STAT.git.

The sequence analysis handles sequential discrete events and behaviors, which can be represented by temporal point processes (TPPs). However, TPP models only occurring events and behaviors. This article explores an efficient method for the negative sequential pattern (NSP) mining to leverage TPP in modeling both frequently occurring and nonoccurring events and behaviors. NSP mining is good at the challenging modeling of nonoccurrences of events and behaviors and their combinations with occurring events, with existing methods built on incorporating various constraints into NSP representations, e.g., simplifying NSP formulations and reducing computational costs. Such constraints restrict the flexibility of NSPs, and nonoccurring behaviors (NOBs) cannot be comprehensively exposed. This article addresses this issue by loosening some inflexible constraints in NSP mining and solves a series of consequent challenges. First, we provide a new definition of negative containment with the set theory according to the loose constraints. Second, an efficient method quickly calculates the supports of negative sequences. Our method only uses the information about the corresponding positive sequential patterns (PSPs) and avoids additional database scans. Finally, a novel and efficient algorithm, NegI-NSP, is proposed to efficiently identify highly valuable NSPs. Theoretical analyses, comparisons, and experiments on four synthetic and two real-life data sets clearly show that NegI-NSP can efficiently discover more useful NOBs.

Heterogeneity across clients in federated learning (FL) usually hinders the optimization convergence and generalization performance when the aggregation of clients' knowledge occurs in the gradient space. For example, clients may differ in terms of data distribution, network latency, input/output space, and/or model architecture, which can easily lead to the misalignment of their local gradients. To improve the tolerance to heterogeneity, we propose a novel federated prototype learning (FedProto) framework in which the clients and server communicate the abstract class prototypes instead of the gradients. FedProto aggregates the local prototypes collected from different clients, and then sends the global prototypes back to all clients to regularize the training of local models. The training on each client aims to minimize the classification error on the local data while keeping the resulting local prototypes sufficiently close to the corresponding global ones. Moreover, we provide a theoretical analysis to the convergence rate of FedProto under non-convex objectives. In experiments, we propose a benchmark setting tailored for heterogeneous FL, with FedProto outperforming several recent FL approaches on multiple datasets.

Text-based games provide an interactive way to study natural language processing.While deep reinforcement learning has shown effectiveness in developing the game playing agent, the low sample efficiency and the large action space remain to be the two major challenges that hinder the DRL from being applied in the real world.In this paper, we address the challenges by introducing world-perceiving modules, which automatically decompose tasks and prune actions by answering questions about the environment.We then propose a two-phase training framework to decouple language learning from reinforcement learning, which further improves the sample efficiency.The experimental results show that the proposed method significantly improves the performance and sample efficiency.Besides, it shows robustness against compound error and limited pre-training data.

Reinforcement learning (RL) is inefficient on long-horizon tasks due to sparse rewards and its policy can be fragile to slightly perturbed environments. We address these challenges via a curriculum of tasks with coupled environments, generated by two policies trained jointly with RL: (1) a co-operative planning policy recursively decomposing a hard task into a coarse-to-fine sub-task tree; and (2) an adversarial policy modifying the environment in each sub-task. They are complementary to acquire more informative feedback for RL: (1) provides dense reward of easier sub-tasks while (2) modifies sub-tasks’ environments to be more challenging and diverse. Conversely, they are trained by RL’s dense feedback on sub-tasks so their generated curriculum keeps adaptive to RL’s progress. The sub-task tree enables an easy-to-hard curriculum for every policy: its top-down construction gradually increases sub-tasks the planner needs to generate, while the adversarial training between the environment and RL follows a bottom-up traversal that starts from a dense sequence of easier sub-tasks allowing more frequent environment changes. We compare EAT-C with RL/planning targeting similar problems and methods with environment generators or adversarial agents. Extensive experiments on diverse tasks demonstrate the advantages of our method on improving RL’s efficiency and generalization.

Unsupervised anomaly detection (AD) is a challenging task in realistic applications. Recently, there is an increasing trend to detect anomalies with deep neural networks (DNN). However, most popular deep AD detectors cannot protect the network from learning contaminated information brought by anomalous data, resulting in unsatisfactory detection performance and overfitting issues. In this work, we identify one reason that hinders most existing DNN-based anomaly detection methods from performing is the wide adoption of the Empirical Risk Minimization (ERM). ERM assumes that the performance of an algorithm on an unknown distribution can be approximated by averaging losses on the known training set. This averaging scheme thus ignores the distinctions between normal and anomalous instances. To break through the limitations of ERM, we propose a novel Diminishing Empirical Risk Minimization (DERM) framework. Specifically, DERM adaptively adjusts the impact of individual losses through a well-devised aggregation strategy. Theoretically, our proposed DERM can directly modify the gradient contribution of each individual loss in the optimization process to suppress the influence of outliers, leading to a robust anomaly detector. Empirically, DERM outperformed the state-of-the-art on the unsupervised AD benchmark consisting of 18 datasets.

In this article, we study the reinforcement learning (RL) for vehicle routing problems (VRPs). Recent works have shown that attention-based RL models outperform recurrent neural network-based methods on these problems in terms of both effectiveness and efficiency. However, existing RL models simply aggregate node embeddings to generate the context embedding without taking into account the dynamic network structures, making them incapable of modeling the state transition and action selection dynamics. In this work, we develop a new attention-based RL model that provides enhanced node embeddings via batch normalization reordering and gate aggregation, as well as dynamic-aware context embedding through an attentive aggregation module on multiple relational structures. We conduct experiments on five types of VRPs: 1) travelling salesman problem (TSP); 2) capacitated VRP (CVRP); 3) split delivery VRP (SDVRP); 4) orienteering problem (OP); and 5) prize collecting TSP (PCTSP). The results show that our model not only outperforms the learning-based baselines but also solves the problems much faster than the traditional baselines. In addition, our model shows improved generalizability when being evaluated in large-scale problems, as well as problems with different data distributions.

We study many-class few-shot (MCFS) problem in both supervised learning and meta-learning settings. Compared to the well-studied many-class many-shot and few-class few-shot problems, the MCFS problem commonly occurs in practical applications but has been rarely studied in previous literature. It brings new challenges of distinguishing between many classes given only a few training samples per class. In this article, we leverage the class hierarchy as a prior knowledge to train a coarse-to-fine classifier that can produce accurate predictions for MCFS problem in both settings. The propose model, “memory-augmented hierarchical-classification network (MahiNet)”, performs coarse-to-fine classification where each coarse class can cover multiple fine classes. Since it is challenging to directly distinguish a variety of fine classes given few-shot data per class, MahiNet starts from learning a classifier over coarse-classes with more training data whose labels are much cheaper to obtain. The coarse classifier reduces the searching range over the fine classes and thus alleviates the challenges from “many classes”. On architecture, MahiNet first deploys a convolutional neural network (CNN) to extract features. It then integrates a memory-augmented attention module and a multi-layer perceptron (MLP) together to produce the probabilities over coarse and fine classes. While the MLP extends the linear classifier, the attention module extends the KNN classifier, both together targeting the “few-shot” problem. We design several training strategies of MahiNet for supervised learning and meta-learning. In addition, we propose two novel benchmark datasets “mcfsImageNet” (as a subset of ImageNet) and “mcfsOmniglot” (re-splitted Omniglot) specially designed for MCFS problem. In experiments, we show that MahiNet outperforms several state-of-the-art models (e.g., prototypical networks and relation networks) on MCFS problems in both supervised learning and meta-learning.

Data similarity (or distance) computation is a fundamental research topic which underpins many high-level applications based on similarity measures in machine learning and data mining. However, in large-scale real-world scenarios, the exact similarity computation has become daunting due to “3V” nature (volume, velocity and variety) of big data. In this case, the hashing techniques have been verified to efficiently conduct similarity estimation in terms of both theory and practice. Currently, MinHash is a popular technique for efficiently estimating the Jaccard similarity of binary sets and furthermore, weighted MinHash is generalized to estimate the generalized Jaccard similarity of weighted sets. This review focuses on categorizing and discussing the existing works of weighted MinHash algorithms. In this review, we mainly categorize the weighted MinHash algorithms into quantization-based approaches, “active index”-based ones and others, and show the evolution and inherent connection of the weighted MinHash algorithms, from the integer weighted MinHash ones to the real-valued weighted MinHash ones. Also, we have developed a Python toolbox for the algorithms, and released it in our github. We experimentally conduct a comprehensive study of the standard MinHash algorithm and the weighted MinHash ones in the similarity estimation error and the information retrieval task.

Multi-view anomaly detection (AD) is a challenging task due to the complicated data distributions across different views. Specifically, there exist two types of anomalies in multi-view distributions: attribute anomaly that exhibits consistent anomalous pattern in each view and class anomaly that exhibits inconsistent traits (e.g., semantic label) across multiple views. Existing methods detect these anomalies in an unsupervised manner with the clustering assumption: normal data share consistent clustering structure across views while anomalous data exhibits inconsistent clusters across views. However, these methods would fail for complex multi-view data distributions where there is no obvious clusters. Moreover, existing models suffer from robustness since they are undermined by anomalies during training time. To get rid of the clustering assumption, we propose a Cross-aligned and Gumbel-refactored AutoEncoders (CGAEs) model to effectively detect two types of multi-view anomalies. In CGAEs, we devise a cross-reconstruction module to detect class anomaly by recovering one view from another view. Class anomalies would lead to high cross-reconstruction loss since they do not have the correct information in one view to generate another. We further design a view-alignment module to detect attribute anomaly by the alignment distance among multiple views in the latent space. Attribute anomalies possess large distances since they are less aligned due to fewer anomalous training instances. To handle the robustness issue, we propose a Gumbel-refactored reconstruction loss to replace the mean square error (MSE) in original autoencoders. The cross entropy loss is calculated between the discreterized input and Gumbel-sampled output, thus disregarding the irrelevant details to achieve model robustness. Experimental results validate the superiority of the proposed CGAEs model on both the benchmark datasets and real world datasets.

Graphs are a powerful and versatile data structure that easily captures real life relationship. Multi-graph Multi-label learning (MGML) is a supervised learning task, which aims to learn a Multi-label classifier to label a set of objects of interest (e.g. image or text) with a bag-of-graphs representation. However, prior techniques on the MGML are developed based on transferring graphs into instances that does not fully utilize the structure information in the learning, and focus on learning the unseen labels only at the bag level. There is no existing work studying how to label the graphs within a bag that is of importance in many applications like image or text annotation. To bridge this gap, in this paper, we present a novel coarse and fine-grained Multi-graph Multi-label (cfMGML) learning framework which directly builds the learning model over the graphs and empowers the label prediction at both the coarse (aka. bag) level and fine-grained (aka. graph in each bag) level. In particular, given a set of labeled multi-graph bags, we design the scoring functions at both graph and bag levels to model the relevance between the label and data using specific graph kernels. Meanwhile, we propose a thresholding rank-loss objective function to rank the labels for the graphs and bags and minimize the hamming-loss simultaneously at one-step, which aims to address the error accumulation issue in traditional rank-loss algorithms. To tackle the non-convex optimization problem, we further develop an effective sub-gradient descent algorithm to handle high-dimensional space computation required in cfMGML. Experiments over various real-world datasets demonstrate cfMGML achieves superior performance than the state-of-arts algorithms.

Traditional network embedding primarily focuses on learning a continuous vector representation for each node, preserving network structure and/or node content information, such that off-the-shelf machine learning algorithms can be easily applied to the vector-format node representations for network analysis. However, the learned continuous vector representations are inefficient for large-scale similarity search, which often involves finding nearest neighbors measured by distance or similarity in a continuous vector space. In this article, we propose a search efficient binary network embedding algorithm called BinaryNE to learn a binary code for each node, by simultaneously modeling node context relations and node attribute relations through a three-layer neural network. BinaryNE learns binary node representations using a stochastic gradient descent-based online learning algorithm. The learned binary encoding not only reduces memory usage to represent each node, but also allows fast bit-wise comparisons to support faster node similarity search than using Euclidean or other distance measures. Extensive experiments and comparisons demonstrate that BinaryNE not only delivers more than 25 times faster search speed, but also provides comparable or better search quality than traditional continuous vector based network embedding methods. The binary codes learned by BinaryNE also render competitive performance on node classification and node clustering tasks. The source code of the BinaryNE algorithm is available at https://github.com/daokunzhang/BinaryNE.

Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial-temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field.

The goal of zero-shot learning (ZSL) is to train a model to classify samples of classes that were not seen during training. To address this challenging task, most ZSL methods relate unseen test classes to seen(training) classes via a pre-defined set of attributes that can describe all classes in the same semantic space, so the knowledge learned on the training classes can be adapted to unseen classes. In this paper, we aim to optimize the attribute space for ZSL by training a propagation mechanism to refine the semantic attributes of each class based on its neighbors and related classes on a graph of classes. We show that the propagated attributes can produce classifiers for zero-shot classes with significantly improved performance in different ZSL settings. The graph of classes is usually free or very cheap to acquire such as WordNet or ImageNet classes. When the graph is not provided, given pre-defined semantic embeddings of the classes, we can learn a mechanism to generate the graph in an end-to-end manner along with the propagation mechanism. However, this graph-aided technique has not been well-explored in the literature. In this paper, we introduce the “attribute propagation network (APNet)”, which is composed of 1) a graph propagation model generating attribute vector for each class and 2) a parameterized nearest neighbor (NN) classifier categorizing an image to the class with the nearest attribute vector to the image's embedding. For better generalization over unseen classes, different from previous methods, we adopt a meta-learning strategy to train the propagation mechanism and the similarity metric for the NN classifier on multiple sub-graphs, each associated with a classification task over a subset of training classes. In experiments with two zero-shot learning settings and five benchmark datasets, APNet achieves either compelling performance or new state-of-the-art results.

In this paper, we study transformers for text-based games. As a promising replacement of recurrent modules in Natural Language Processing (NLP) tasks, the transformer architecture could be treated as a powerful state representation generator for reinforcement learning. However, the vanilla transformer is neither effective nor efficient to learn with a huge amount of weight parameters. Unlike existing research that encodes states using LSTMs or GRUs, we develop a novel lightweight transformer-based representation generator featured with reordered layer normalization, weight sharing and block-wise aggregation. The experimental results show that our proposed model not only solves single games with much fewer interactions, but also achieves better generalization on a set of unseen games. Furthermore, our model outperforms state-of-the-art agents in a variety of man-made games.

Many graph embedding approaches have been proposed for knowledge graph completion via link prediction.Among those, translating embedding approaches enjoy the advantages of lightweight structure, high efficiency and great interpretability.Especially when extended to complex vector space, they show the capability in handling various relation patterns including symmetry, antisymmetry, inversion and composition.However, previous translating embedding approaches defined in complex vector space suffer from two main issues: 1) representing and modeling capacities of the model are limited by the translation function with rigorous multiplication of two complex numbers; and 2) embedding ambiguity caused by one-to-many relations is not explicitly alleviated.In this paper, we propose a relation-adaptive translation function built upon a novel weighted product in complex space, where the weights are learnable, relation-specific and independent to embedding size.The translation function only requires eight more scalar parameters each relation, but improves expressive power and alleviates embedding ambiguity problem.Based on the function, we then present our Relation-adaptive translating Embedding (RatE) approach to score each graph triple.Moreover, a novel negative sampling method is proposed to utilize both prior knowledge and self-adversarial learning for effective optimization.Experiments verify RatE achieves state-of-the-art performance on four link prediction benchmarks.

Wrong labeling problem and long-tail relations are two main challenges caused by distant supervision in relation extraction.Recent works alleviate the wrong labeling by selective attention via multi-instance learning, but cannot well handle long-tail relations even if hierarchies of the relations are introduced to share knowledge.In this work, we propose a novel neural network, Collaborating Relation-augmented Attention (CoRA), to handle both the wrong labeling and long-tail relations.Particularly, we first propose relation-augmented attention network as base model.It operates on sentence bag with a sentence-to-relation attention to minimize the effect of wrong labeling.Then, facilitated by the proposed base model, we introduce collaborating relation features shared among relations in the hierarchies to promote the relation-augmenting process and balance the training data for long-tail relations.Besides the main training objective to predict the relation of a sentence bag, an auxiliary objective is utilized to guide the relationaugmenting process for a more accurate bag-level representation.In the experiments on the popular benchmark dataset NYT, the proposed CoRA improves the prior state-of-the-art performance by a large margin in terms of Precision@N, AUC and [email protected] analyses verify its superior capability in handling long-tail relations in contrast to the competitors.

Electronic health records (EHRs) are longitudinal records of a patient's interactions with healthcare systems. A patient's EHR data is organized as a three-level hierarchy from top to bottom: patient journey - all the experiences of diagnoses and treatments over a period of time; individual visit - a set of medical codes in a particular visit; and medical code - a specific record in the form of medical codes. As EHRs begin to amass in millions, the potential benefits, which these data might hold for medical research and medical outcome prediction, are staggering - including, for example, predicting future admissions to hospitals, diagnosing illnesses or determining the efficacy of medical treatments. Each of these analytics tasks requires a domain knowledge extraction method to transform the hierarchical patient journey into a vector representation for further prediction procedure. The representations should embed a sequence of visits and a set of medical codes with a specific timestamp, which are crucial to any downstream prediction tasks. Hence, expressively powerful representations are appealing to boost learning performance. To this end, we propose a novel self-attention mechanism that captures the contextual dependency and temporal relationships within a patient's healthcare journey. An end-to-end bidirectional temporal encoder network (BiteNet) then learns representations of the patient's journeys, based solely on the proposed attention mechanism. We have evaluated the effectiveness of our methods on two supervised prediction and two unsupervised clustering tasks with a real-world EHR dataset. The empirical results demonstrate the proposed BiteNet model produces higher-quality representations than state-of-the-art baseline methods.

Many algorithms for Knowledge-Based Question Answering (KBQA) depend on semantic parsing, which translates a question to its logical form. When only weak supervision is provided, it is usually necessary to search valid logical forms for model training. However, a complex question typically involves a huge search space, which creates two main problems: 1) the solutions limited by computation time and memory usually reduce the success rate of the search, and 2) spurious logical forms in the search results degrade the quality of training data. These two problems lead to a poorly-trained semantic parsing model. In this work, we propose an effective search method for weakly supervised KBQA based on operator prediction for questions. With search space constrained by predicted operators, sufficient search paths can be explored, more valid logical forms can be derived, and operators possibly causing spurious logical forms can be avoided. As a result, a larger proportion of questions in a weakly supervised training set are equipped with logical forms, and fewer spurious logical forms are generated. Such high-quality training data directly contributes to a better semantic parsing model. Experimental results on one of the largest KBQA datasets (i.e., CSQA) verify the effectiveness of our approach and deliver a new state-of-the-art performance.

Modeling multivariate time series has long been a subject that has attracted researchers from a diverse range of fields including economics, finance, and traffic. A basic assumption behind multivariate time series forecasting is that its variables depend on one another but, upon looking closely, it is fair to say that existing methods fail to fully exploit latent spatial dependencies between pairs of variables. In recent years, meanwhile, graph neural networks (GNNs) have shown high capability in handling relational dependencies. GNNs require well-defined graph structures for information propagation which means they cannot be applied directly for multivariate time series where the dependencies are not known in advance. In this paper, we propose a general graph neural network framework designed specifically for multivariate time series data. Our approach automatically extracts the uni-directed relations among variables through a graph learning module, into which external knowledge like variable attributes can be easily integrated. A novel mix-hop propagation layer and a dilated inception layer are further proposed to capture the spatial and temporal dependencies within the time series. The graph learning, graph convolution, and temporal convolution modules are jointly learned in an end-to-end framework. Experimental results show that our proposed model outperforms the state-of-the-art baseline methods on 3 of 4 benchmark datasets and achieves on-par performance with other approaches on two traffic datasets which provide extra structural information.

We study reinforcement learning (RL) for text-based games, which are interactive simulations in the context of natural language. While different methods have been developed to represent the environment information and language actions, existing RL agents are not empowered with any reasoning capabilities to deal with textual games. In this work, we aim to conduct explicit reasoning with knowledge graphs for decision making, so that the actions of an agent are generated and supported by an interpretable inference procedure. We propose a stacked hierarchical attention mechanism to construct an explicit representation of the reasoning process by exploiting the structure of the knowledge graph. We extensively evaluate our method on a number of man-made benchmark games, and the experimental results demonstrate that our method performs better than existing text-based agents.

Numerous deep reinforcement learning agents have been proposed, and each of them has its strengths and flaws. In this work, we present a Cooperative Heterogeneous Deep Reinforcement Learning (CHDRL) framework that can learn a policy by integrating the advantages of heterogeneous agents. Specifically, we propose a cooperative learning framework that classifies heterogeneous agents into two classes: global agents and local agents. Global agents are off-policy agents that can utilize experiences from the other agents. Local agents are either on-policy agents or population-based evolutionary algorithms (EAs) agents that can explore the local area effectively. We employ global agents, which are sample-efficient, to guide the learning of local agents so that local agents can benefit from the sample-efficient agents and simultaneously maintain their advantages, e.g., stability. Global agents also benefit from effective local searches. Experimental studies on a range of continuous control tasks from the Mujoco benchmark show that CHDRL achieves better performance compared with state-of-the-art baselines.

With the widespread use of information technologies, information networks are becoming increasingly popular to capture complex relationships across various disciplines, such as social networks, citation networks, telecommunication networks, and biological networks. Analyzing these networks sheds light on different aspects of social life such as the structure of societies, information diffusion, and communication patterns. In reality, however, the large scale of information networks often makes network analytic tasks computationally expensive or intractable. Network representation learning has been recently proposed as a new learning paradigm to embed network vertices into a low-dimensional vector space, by preserving network topology structure, vertex content, and other side information. This facilitates the original network to be easily handled in the new vector space for further analysis. In this survey, we perform a comprehensive review of the current literature on network representation learning in the data mining and machine learning field. We propose new taxonomies to categorize and summarize the state-of-the-art network representation learning techniques according to the underlying learning mechanisms, the network information intended to preserve, as well as the algorithmic designs and methodologies. We summarize evaluation protocols used for validating network representation learning including published benchmark datasets, evaluation methods, and open source algorithms. We also perform empirical studies to compare the performance of representative algorithms on common datasets, and analyze their computational complexity. Finally, we suggest promising research directions to facilitate future study.

Graph embedding aims to transfer a graph into vectors to facilitate subsequent graph-analytics tasks like link prediction and graph clustering. Most approaches on graph embedding focus on preserving the graph structure or minimizing the reconstruction errors for graph data. They have mostly overlooked the embedding distribution of the latent codes, which unfortunately may lead to inferior representation in many cases. In this article, we present a novel adversarially regularized framework for graph embedding. By employing the graph convolutional network as an encoder, our framework embeds the topological information and node content into a vector representation, from which a graph decoder is further built to reconstruct the input graph. The adversarial training principle is applied to enforce our latent codes to match a prior Gaussian or uniform distribution. Based on this framework, we derive two variants of the adversarial models, the adversarially regularized graph autoencoder (ARGA) and its variational version, and adversarially regularized variational graph autoencoder (ARVGA), to learn the graph embedding effectively. We also exploit other potential variations of ARGA and ARVGA to get a deeper understanding of our designs. Experimental results that compared 12 algorithms for link prediction and 20 algorithms for graph clustering validate our solutions.

Matrix factorization has been widely applied to various applications. With the fast development of storage and internet technologies, we have been witnessing a rapid increase of data. In this paper, we propose new algorithms for matrix factorization with the emphasis on efficiency. In addition, most existing methods of matrix factorization only consider a general smooth least square loss. Differently, many real-world applications have distinctive characteristics. As a result, different losses should be used accordingly. Therefore, it is beneficial to design new matrix factorization algorithms that are able to deal with both smooth and non-smooth losses. To this end, one needs to analyze the characteristics of target data and use the most appropriate loss based on the analysis. We particularly study two representative cases of low-rank matrix recovery, i.e., collaborative filtering for recommendation and high dynamic range imaging. To solve these two problems, we respectively propose a stage-wise matrix factorization algorithm by exploiting manifold optimization techniques. From our theoretical analysis, they are both are provably guaranteed to converge to a stationary point. Extensive experiments on recommender systems and high dynamic range imaging demonstrate the satisfactory performance and efficiency of our proposed method on large-scale real data.

Graphs are commonly used to represent objects, such as images and text, for pattern classification. In a dynamic world, an object may continuously evolve over time, and so does the graph extracted from the underlying object. These changes in graph structure with respect to the temporal order present a new representation of the graph, in which an object corresponds to a set of time-variant graphs. In this paper, we formulate a novel time-variant graph classification task and propose a new graph feature, called a graph-shapelet pattern, for learning and classifying time-variant graphs. Graph-shapelet patterns are compact and discriminative graph transformation subsequences. A graph-shapelet pattern can be regarded as a graphical extension of a shapelet-a class of discriminative features designed for vector-based temporal data classification. To discover graph-shapelet patterns, we propose to convert a time-variant graph sequence into time-series data and use the discovered shapelets to find graph transformation subsequences as graph-shapelet patterns. By converting each graph-shapelet pattern into a unique tokenized graph transformation sequence, we can measure the similarity between two graph-shapelet patterns and therefore classify time-variant graphs. Experiments on both synthetic and real-world data demonstrate the superior performance of the proposed algorithms.

When users in online social networks make a decision, they are often affected by their neighbors. Social recommendation models utilize social information to reveal the impact of neighbors on user preferences, and this impact is often described by the linear superposition of neighbor preferences or by global trust propagation. Further exploration needs to be undertaken to determine whether the influence pattern of other users from online interaction behaviors is adequately described. In this paper, we introduce evolutionary opinion dynamics from the field of statistical physics into recommender systems, characterizing the impact of other users. We propose an opinion dynamic model by evolutionary game theory. To describe online user interactions, we define the strategies during an interaction between two users, and present the payoff for each strategy in terms of errors of estimated ratings. Therefore, user behaviors are associated with their preferences and ratings. In addition, we measure user influence according to their topological roles in the social network. We incorporate evolutionary opinion dynamics and user influence into the recommendation framework for the prediction of unknown ratings. Experiment results on two real-world datasets demonstrate that our method outperforms state-of the-art models in terms of accuracy, and it also performs well for cold-start users. Our method reduces the divergence of user preferences, in accordance with online opinion interactions. Furthermore, our method has approximate computational complexity with matrix factorization, and results in less computation than state-of-the-art models. Our method is quite general, and indicates that studies in social physics, statistics, and other research fields may be involved in recommendation to improve the performance.

A variety of machine learning applications expect to achieve rapid learning from a limited number of labeled data. However, the success of most current models is the result of heavy training on big data. Meta-learning addresses this problem by extracting common knowledge across different tasks that can be quickly adapted to new tasks. However, they do not fully explore weakly-supervised information, which is usually free or cheap to collect. In this paper, we show that weakly-labeled data can significantly improve the performance of meta-learning on few-shot classification. We propose prototype propagation network (PPN) trained on few-shot tasks together with data annotated by coarse-label. Given a category graph of the targeted fine-classes and some weakly-labeled coarse-classes, PPN learns an attention mechanism which propagates the prototype of one class to another on the graph, so that the K-nearest neighbor (KNN) classifier defined on the propagated prototypes results in high accuracy across different few-shot tasks. The training tasks are generated by subgraph sampling, and the training objective is obtained by accumulating the level-wise classification loss on the subgraph. On two benchmarks, PPN significantly outperforms most recent few-shot learning methods in different settings, even when they are also allowed to train on weakly-labeled data.

Graph clustering is a fundamental task which discovers communities or groups in networks. Recent studies have mostly focused on developing deep learning approaches to learn a compact graph embedding, upon which classic clustering methods like k-means or spectral clustering algorithms are applied. These two-step frameworks are difficult to manipulate and usually lead to suboptimal performance, mainly because the graph embedding is not goal-directed, i.e., designed for the specific clustering task. In this paper, we propose a goal-directed deep learning approach, Deep Attentional Embedded Graph Clustering (DAEGC for short). Our method focuses on attributed graphs to sufficiently explore the two sides of information in graphs. By employing an attention network to capture the importance of the neighboring nodes to a target node, our DAEGC algorithm encodes the topological structure and node content in a graph to a compact representation, on which an inner product decoder is trained to reconstruct the graph structure. Furthermore, soft labels from the graph embedding itself are generated to supervise a self-training graph clustering process, which iteratively refines the clustering results. The self-training process is jointly learned and optimized with the graph embedding in a unified framework, to mutually benefit both components. Experimental results compared with state-of-the-art algorithms demonstrate the superiority of our method.

Spatial-temporal graph modeling is an important task to analyze the spatial relations and temporal trends of components in a system. Existing approaches mostly capture the spatial dependency on a fixed graph structure, assuming that the underlying relation between entities is pre-determined. However, the explicit graph structure (relation) does not necessarily reflect the true dependency and genuine relation may be missing due to the incomplete connections in the data. Furthermore, existing methods are ineffective to capture the temporal trends as the RNNs or CNNs employed in these methods cannot capture long-range temporal sequences. To overcome these limitations, we propose in this paper a novel graph neural network architecture, {Graph WaveNet}, for spatial-temporal graph modeling. By developing a novel adaptive dependency matrix and learn it through node embedding, our model can precisely capture the hidden spatial dependency in the data. With a stacked dilated 1D convolution component whose receptive field grows exponentially as the number of layers increases, Graph WaveNet is able to handle very long sequences. These two components are integrated seamlessly in a unified framework and the whole framework is learned in an end-to-end manner. Experimental results on two public traffic network datasets, METR-LA and PEMS-BAY, demonstrate the superior performance of our algorithm.

Meta-learning extracts the common knowledge from learning different tasks and uses it for unseen tasks. It can significantly improve tasks that suffer from insufficient training data, e.g., few-shot learning. In most meta-learning methods, tasks are implicitly related by sharing parameters or optimizer. In this paper, we show that a meta-learner that explicitly relates tasks on a graph describing the relations of their output dimensions (e.g., classes) can significantly improve few-shot learning. The graph’s structure is usually free or cheap to obtain but has rarely been explored in previous works. We develop a novel meta-learner of this type for prototype based classification, in which a prototype is generated for each class, such that the nearest neighbor search among the prototypes produces an accurate classification. The meta-learner, called “Gated Propagation Network (GPN)”, learns to propagate messages between prototypes of different classes on the graph, so that learning the prototype of each class benefits from the data of other related classes. In GPN, an attention mechanism aggregates messages from neighboring classes of each class, with a gate choosing between the aggregated message and the message from the class itself. We train GPN on a sequence of tasks from many-shot to few-shot generated by subgraph sampling. During training, it is able to reuse and update previously achieved prototypes from the memory in a life-long learning cycle. In experiments, under different training-test discrepancy and test task generation settings, GPN outperforms recent meta-learning methods on two benchmark datasets. Code of GPN is publicly available at: https://github.com/liulu112601/Gated-Propagation-Net.

Time series has been a popular research topic over the past decade. Salient subsequences of time series that can benefit the learning task, e.g., classification or clustering, are called shapelets. Shapelet-based time series learning extracts these types of salient subsequences with highly informative features from a time series. Most existing methods for shapelet discovery must scan a large pool of candidate subsequences, which is a time-consuming process. A recent work, [1] , uses regression learning to discover shapelets in a time series; however, it only considers learning shapelets from labeled time series data. This paper proposes an Unsupervised Salient Subsequence Learning (USSL) model that discovers shapelets without the effort of labeling. We developed this new learning function by integrating the strengths of shapelet learning, shapelet regularization, spectral analysis and pseudo-label to simultaneously and automatically learn shapelets to help clustering unlabeled time series better. The optimization model is iteratively solved via a coordinate descent algorithm. Experiments show that our USSL can learn meaningful shapelets, with promising results on real-world and synthetic data that surpass current state-of-the-art unsupervised time series learning methods.

There is an emerging trend to leverage noisy image datasets in many visual recognition tasks. However, the label noise among datasets severely degenerates the performance of deep learning approaches. Recently, one mainstream is to introduce the latent label to handle label noise, which has shown promising improvement in the network designs. Nevertheless, the mismatch between latent labels and noisy labels still affects the predictions in such methods. To address this issue, we propose a probabilistic model, which explicitly introduces an extra variable to represent the trustworthiness of noisy labels, termed as the quality variable. Our key idea is to identify the mismatch between the latent and noisy labels by embedding the quality variables into different subspaces, which effectively minimizes the influence of label noise. At the same time, reliable labels are still able to be applied for training. To instantiate the model, we further propose a Contrastive-Additive Noise network (CAN), which consists of two important layers: (1) the contrastive layer that estimates the quality variable in the embedding space to reduce the influence of noisy labels; and (2) the additive layer that aggregates the prior prediction and noisy labels as the posterior to train the classifier. Moreover, to tackle the challenges in optimization, we deduce an SGD algorithm with the reparameterization tricks, which makes our method scalable to big data.We validate the proposed method on a range of noisy image datasets. Comprehensive results have demonstrated that CAN outperforms the state-of-the-art deep learning approaches.

Networked data are common in domains where instances are characterized by both feature values and inter-dependency relationships. Finding cluster structures for networked instances and discovering representative features for each cluster represent a special co-clustering task usefully for many real-world applications, such as automatic categorization of scientific publications and finding representative key-words for each cluster. To date, although co-clustering has been commonly used for finding clusters for both instances and features, all existing methods are focused on instance-feature values, without leveraging valuable topology relationships between instances to help boost co-clustering performance. In this paper, we propose CFOND, a consensus factorization based framework for co-clustering networked data. We argue that feature values and linkages provide useful information from different perspectives, but they are not always consistent and therefore need to be carefully aligned for best clustering results. In the paper, we advocate a consensus factorization principle, which simultaneously factorizes information from three aspects: network topology structures, instance-feature content relationships, and feature-feature correlations. The consensus factorization ensures that the final cluster structures are consistent across information from the three aspects with minimum errors. Experiments on real-life networks validate the performance of our algorithm.

Min-Hash is a popular technique for efficiently estimating the Jaccard similarity of binary sets. Consistent Weighted Sampling (CWS) generalizes the Min-Hash scheme to sketch weighted sets and has drawn increasing interest from the community. Due to its constant-time complexity independent of the values of the weights, Improved CWS (ICWS) is considered as the state-of-the-art CWS algorithm. In this paper, we revisit ICWS and analyze its underlying mechanism to show that there actually exists dependence between the two components of the hash-code produced by ICWS, which violates the condition of independence. To remedy the problem, we propose an Improved ICWS (I2CWS) algorithm which not only shares the same theoretical computational complexity as ICWS but also abides by the required conditions of the CWS scheme. The experimental results on a number of synthetic data sets and real-world text data sets demonstrate that our I2CWS algorithm can estimate the Jaccard similarity more accurately, and also competes with or outperforms the compared methods, including ICWS, in classification and top-K retrieval, after relieving the underlying dependence.

Recurrent neural nets (RNN) and convolutional neural nets (CNN) are widely used on NLP tasks to capture the long-term and local dependencies, respectively. Attention mechanisms have recently attracted enormous interest due to their highly parallelizable computation, significantly less training time, and flexibility in modeling dependencies. We propose a novel attention mechanism in which the attention between elements from input sequence(s) is directional and multi-dimensional (i.e., feature-wise). A light-weight neural net, "Directional Self-Attention Network (DiSAN)," is then proposed to learn sentence embedding, based solely on the proposed attention without any RNN/CNN structure. DiSAN is only composed of a directional self-attention with temporal order encoded, followed by a multi-dimensional attention that compresses the sequence into a vector representation. Despite its simple form, DiSAN outperforms complicated RNN models on both prediction quality and time efficiency. It achieves the best test accuracy among all sentence encoding methods and improves the most recent best result by 1.02% on the Stanford Natural Language Inference (SNLI) dataset, and shows state-of-the-art test accuracy on the Stanford Sentiment Treebank (SST), Multi-Genre natural language inference (MultiNLI), Sentences Involving Compositional Knowledge (SICK), Customer Review, MPQA, TREC question-type classification and Subjectivity (SUBJ) datasets.

Attributed network embedding enables joint representation learning of node links and attributes. Existing attributed network embedding models are designed in continuous Euclidean spaces which often introduce data redundancy and impose challenges to storage and computation costs. To this end, we present a Binarized Attributed Network Embedding model (BANE for short) to learn binary node representation. Specifically, we define a new Weisfeiler-Lehman proximity matrix to capture data dependence between node links and attributes by aggregating the information of node attributes and links from neighboring nodes to a given target node in a layer-wise manner. Based on the Weisfeiler-Lehman proximity matrix, we formulate a new Weisfiler-Lehman matrix factorization learning function under the binary node representation constraint. The learning problem is a mixed integer optimization and an efficient cyclic coordinate descent (CCD) algorithm is used as the solution. Node classification and link prediction experiments on real-world datasets show that the proposed BANE model outperforms the state-of-the-art network embedding methods.

Attributed network embedding aims to learn low-dimensional vector representations for nodes in a network, where each node contains rich attributes/features describing node content. Because network topology structure and node attributes often exhibit high correlation, incorporating node attribute proximity into network embedding is beneficial for learning good vector representations. In reality, large-scale networks often have incomplete/missing node content or linkages, yet existing attributed network embedding algorithms all operate under the assumption that networks are complete. Thus, their performance is vulnerable to missing data and suffers from poor scalability. In this paper, we propose a Scalable Incomplete Network Embedding (SINE) algorithm for learning node representations from incomplete graphs. SINE formulates a probabilistic learning framework that separately models pairs of node-context and node-attribute relationships. Different from existing attributed network embedding algorithms, SINE provides greater flexibility to make the best of useful information and mitigate negative effects of missing information on representation learning. A stochastic gradient descent based online algorithm is derived to learn node representations, allowing SINE to scale up to large-scale networks with high learning efficiency. We evaluate the effectiveness and efficiency of SINE through extensive experiments on real-world networks. Experimental results confirm that SINE outperforms state-of-the-art baselines in various tasks, including node classification, node clustering, and link prediction, under settings with missing links and node attributes. SINE is also shown to be scalable and efficient on large-scale networks with millions of nodes/edges and high-dimensional node features. The source code of this paper is available at https://github.com/daokunzhang/SINE.

Graph embedding is an effective method to represent graph data in a low dimensional space for graph analytics. Most existing embedding algorithms typically focus on preserving the topological structure or minimizing the reconstruction errors of graph data, but they have mostly ignored the data distribution of the latent codes from the graphs, which often results in inferior embedding in real-world graph data. In this paper, we propose a novel adversarial graph embedding framework for graph data. The framework encodes the topological structure and node content in a graph to a compact representation, on which a decoder is trained to reconstruct the graph structure. Furthermore, the latent representation is enforced to match a prior distribution via an adversarial training scheme. To learn a robust embedding, two variants of adversarial approaches, adversarially regularized graph autoencoder (ARGA) and adversarially regularized variational graph autoencoder (ARVGA), are developed. Experimental studies on real-world graphs validate our design and demonstrate that our algorithms outperform baselines by a wide margin in link prediction, graph clustering, and graph visualization tasks.

Many natural language processing tasks solely rely on sparse dependencies between a few tokens in a sentence. Soft attention mechanisms show promising performance in modeling local/global dependencies by soft probabilities between every two tokens, but they are not effective and efficient when applied to long sentences. By contrast, hard attention mechanisms directly select a subset of tokens but are difficult and inefficient to train due to their combinatorial nature. In this paper, we integrate both soft and hard attention into one context fusion model, "reinforced self-attention (ReSA)", for the mutual benefit of each other. In ReSA, a hard attention trims a sequence for a soft self-attention to process, while the soft attention feeds reward signals back to facilitate the training of the hard one. For this purpose, we develop a novel hard attention called "reinforced sequence sampling (RSS)", selecting tokens in parallel and trained via policy gradient. Using two RSS modules, ReSA efficiently extracts the sparse dependencies between each pair of selected tokens. We finally propose an RNN/CNN-free sentence-encoding model, "reinforced self-attention network (ReSAN)", solely based on ReSA. It achieves state-of-the-art performance on both the Stanford Natural Language Inference (SNLI) and the Sentences Involving Compositional Knowledge (SICK) datasets.

Attributed network embedding aims to learn a low-dimensional representation for each node of a network, considering both attributes and structure information of the node. However, the learning based methods usually involve substantial cost in time, which makes them impractical without the help of a powerful workhorse. In this paper, we propose a simple yet effective algorithm, named NetHash, to solve this problem only with moderate computing capacity. NetHash employs the randomized hashing technique to encode shallow trees, each of which is rooted at a node of the network. The main idea is to efficiently encode both attributes and structure information of each node by recursively sketching the corresponding rooted tree from bottom (i.e., the predefined highest-order neighboring nodes) to top (i.e., the root node), and particularly, to preserve as much information closer to the root node as possible. Our extensive experimental results show that the proposed algorithm, which does not need learning, runs significantly faster than the state-of-the-art learning-based network embedding methods while achieving competitive or even better performance in accuracy.

Analyzing accumulated data has recently attracted huge attention for its ability to generate values by identifying useful information and providing an edge in global business competition. However, heterogeneous data and imbalanced class distribution present two major challenges to machine learning with real-world business data. Traditional machine learning algorithms can typically only be applied to standard data sets, which are normally homogeneous and balanced. These algorithms narrow complex data into a homogeneous, a balanced data space an inefficient process that requires a significant amount of pre-processing. In this paper, we focus on an efficient solution to the challenges with heterogeneous and imbalanced data sets that does not require pre-processing. Our approach comprises a novel, unified, end-to-end cost-sensitive hybrid neural network that learns real-world heterogeneous data via a parallel network architecture. A specifically-designed cost-sensitive matrix then automatically generates a robust model for learning minority classifications. And the parameters of both the cost-sensitive matrix and the hybrid neural network are alternately but jointly optimized during training. The results of comparative experiments on six real-world data sets reflecting actual business cases, including insurance fraud detection and mobile customer demographics, indicate that the proposed approach demonstrates superior performance over baseline procedures.

Existing graph classification usually relies on an exhaustive enumeration of substructure patterns, where the number of substructures expands exponentially w.r.t. with the size of the graph set. Recently, the Weisfeiler-Lehman (WL) graph kernel has achieved the best performance in terms of both accuracy and efficiency among state-of-the-art methods. However, it is still time-consuming, especially for large-scale graph classification tasks. In this paper, we present a K-Ary Tree based Hashing (KATH) algorithm, which is able to obtain competitive accuracy with a very fast runtime. The main idea of KATH is to construct a traversal table to quickly approximate the subtree patterns in WL using K-ary trees. Based on the traversal table, KATH employs a recursive indexing process that performs only r times of matrix indexing to generate all (r - 1)-depth K-ary trees, where the leaf node labels of a tree can uniquely specify the pattern. After that, the MinHash scheme is used to fingerprint the acquired subtree patterns for a graph. Our experimental results on both real world and synthetic data sets show that KATH runs significantly faster than state-of-the-art methods while achieving competitive or better accuracy.

Multi-instance learning (MIL) is a useful tool for tackling labeling ambiguity in learning because it allows a bag of instances to share one label. Bag mapping transforms a bag into a single instance in a new space via instance selection and has drawn significant attention recently. To date, most existing work is based on the original space, using all instances inside each bag for bag mapping, and the selected instances are not directly tied to an MIL objective. As a result, it is difficult to guarantee the distinguishing capacity of the selected instances in the new bag mapping space. In this paper, we propose a discriminative mapping approach for multi-instance learning (MILDM) that aims to identify the best instances to directly distinguish bags in the new mapping space. Accordingly, each instance bag can be mapped using the selected instances to a new feature space, and hence any generic learning algorithm, such as an instance-based learning algorithm, can be used to derive learning models for multi-instance classification. Experiments and comparisons on eight different types of real-world learning tasks (including 14 data sets) demonstrate that MILDM outperforms the state-of-the-art bag mapping multi-instance learning approaches. Results also confirm that MILDM achieves balanced performance between runtime efficiency and classification effectiveness.

Spectral clustering (SC) has been widely applied to various computer vision tasks, where the key is to construct a robust affinity matrix for data partitioning. With the increase in visual features, conventional SC methods are facing two challenges: 1) how to effectively generate an affinity matrix based on multiple features? and 2) how to deal with high-dimensional visual features which could be redundant? To address these issues mentioned earlier, we present a new approach to: 1) learn a robust affinity matrix using multiple features, allowing us to simultaneously determine optimal weights for each feature; and 2) decide a set of optimal projection matrixes, one for each feature, that decide the lower dimensional space, as well as the optimal affinity weight of each data pair in the lower dimensional space. There are two major advantages of our new approach over the existing clustering techniques. First, our approach assigns affinity weights for data points on a per-data-pair basis. The learning procedure avoids the explicit specification of the size of the neighborhood in the affinity matrix, and the bandwidth parameter required to compute the Gaussian kernel, both of which are sensitive and yet difficult to determine beforehand. Second, the affinity weights are based on the distances in a lower dimensional space, while the low-dimensional space is inferred according to the optimized affinity weights. Both variables are jointly optimized so as to leverage mutual benefits. The experimental results outperform the compared alternatives, which indicate that the proposed method is effective in simultaneously learning the affinity graph and feature fusion, resulting in better clustering results.

Many applications involve objects containing structure and rich content information, each describing different feature aspects of the object. Graph learning and classification is a common tool for handling such objects. To date, existing graph classification has been limited to the single-graph setting with each object being represented as one graph from a single structure-view. This inherently limits its use to the classification of complicated objects containing complex structures and uncertain labels. In this paper, we advance graph classification to handle multigraph learning for complicated objects from multiple structure views, where each object is represented as a bag containing several graphs and the label is only available for each graph bag but not individual graphs inside the bag. To learn such graph classification models, we propose a multistructure-view bag constrained learning (MSVBL) algorithm, which aims to explore substructure features across multiple structure views for learning. By enabling joint regularization across multiple structure views and enforcing labeling constraints at the bag and graph levels, MSVBL is able to discover the most effective substructure features across all structure views. Experiments and comparisons on real-world data sets validate and demonstrate the superior performance of MSVBL in representing complicated objects as multigraph for classification, e.g., MSVBL outperforms the state-of-the-art multiview graph classification and multiview multi-instance learning approaches.

© 2017, Copyright is with the authors. In spatial crowdsourcing, spatial tasks are outsourced to a set of workers in proximity of the task locations for efficient assignment. It usually requires workers to disclose their locations, which inevitably raises security concerns about the privacy of the workers’ locations. In this paper, we propose a secure SC framework based on encryption, which ensures that workers’ location information is never released to any party, yet the system can still assign tasks to workers situated in proximity of each task’s location. We solve the challenge of assigning tasks based on encrypted data using homomorphic encryption. Moreover, to overcome the efficiency issue, we propose a novel secure indexing technique with a newly devised SKD-tree to index encrypted worker locations. Experiments on real-world data evaluate various aspects of the performance of the proposed SC platform.

Review rating prediction is an important research topic. The problem was approached from either the perspective of recommender systems (RS) or that of sentiment analysis (SA). Recent SA research using deep neural networks (DNNs) has realized the importance of user and product interaction for better interpreting the sentiment of reviews. However, the complexity of DNN models in terms of the scale of parameters is very high, and the performance is not always satisfying especially when user-product interaction is sparse. In this paper, we propose a simple, extensible RS-based model, called Text-driven Latent Factor Model (TLFM), to capture the semantics of reviews, user preferences and product characteristics by jointly optimizing two components, a user-specific LFM and a product-specific LFM, each of which decomposes text into a specific low-dimension representation. Furthermore, we address the cold-start issue by developing a novel Pairwise Rating Comparison strategy (PRC), which utilizes the difference between ratings on common user/product as supplementary information to calibrate parameter estimation. Experiments conducted on IMDB and Yelp datasets validate the advantage of our approach over state-of-the-art baseline methods.

This paper addresses social network embedding, which aims to embed social network nodes, including user profile information, into a latent low-dimensional space. Most of the existing works on network embedding only consider network structure, but ignore user-generated content that could be potentially helpful in learning a better joint network representation. Different from rich node content in citation networks, user profile information in social networks is useful but noisy, sparse, and incomplete. To properly utilize this information, we propose a new algorithm called User Profile Preserving Social Network Embedding (UPP-SNE), which incorporates user profile with network structure to jointly learn a vector representation of a social network. The theme of UPP-SNE is to embed user profile information via a nonlinear mapping into a consistent subspace, where network structure is seamlessly encoded to jointly learn informative node representations. Extensive experiments on four real-world social networks show that compared to state-of-the-art baselines, our method learns better social network representations and achieves substantial performance gains in node classification and clustering tasks.

Min-Hash, which is widely used for efficiently estimating similarities of bag-of-words represented data, plays an increasingly important role in the era of big data. It has been extended to deal with real-value weighted sets -- Improved Consistent Weighted Sampling (ICWS) is considered as the state-of-the-art for this problem. In this paper, we propose a Practical CWS (PCWS) algorithm. We first transform the original form of ICWS into an equivalent expression, based on which we find some interesting properties that inspire us to make the ICWS algorithm simpler and more efficient in both space and time complexities. PCWS is not only mathematically equivalent to ICWS and preserves the same theoretical properties, but also saves 20% memory footprint and substantial computational cost compared to ICWS. The experimental results on a number of real-world text data sets demonstrate that PCWS obtains the same (even better) classification and retrieval performance as ICWS with 1/5~1/3 reduced empirical runtime.

Frequently recurring RNA structural motifs play important roles in RNA folding process and interaction with other molecules. Traditional index-based and shape-based schemas are useful in modeling RNA secondary structures but ignore the structural discrepancy of individual RNA family member. Further, the in-depth analysis of underlying substructure pattern is insufficient due to varied and unnormalized substructure data. This prevents us from understanding RNAs functions and their inherent synergistic regulation networks. This article thus proposes a novel labeled graph-based algorithm RnaGraph to uncover frequently RNA substructure patterns. Attribute data and graph data are combined to characterize diverse substructures and their correlations, respectively. Further, a top-k graph pattern mining algorithm is developed to extract interesting substructure motifs by integrating frequency and similarity. The experimental results show that our methods assist in not only modelling complex RNA secondary structures but also identifying hidden but interesting RNA substructure patterns.

Multitask learning (MTL) is commonly used for jointly optimizing multiple learning tasks. To date, all existing MTL methods have been designed for tasks with feature-vector represented instances, but cannot be applied to structure data, such as graphs. More importantly, when carrying out MTL, existing methods mainly focus on exploring overall commonality or disparity between tasks for learning, but cannot explicitly capture task relationships in the feature space, so they are unable to answer important questions, such as what exactly is shared between tasks and what is the uniqueness of one task differing from others? In this paper, we formulate a new multitask graph learning problem, and propose a task sensitive feature exploration and learning algorithm for multitask graph classification. Because graphs do not have features available, we advocate a task sensitive feature exploration and learning paradigm to jointly discover discriminative subgraph features across different tasks. In addition, a feature learning process is carried out to categorize each subgraph feature into one of three categories: (1) common feature; (2) task auxiliary feature; and (3) task specific feature, indicating whether the feature is shared by all tasks, by a subset of tasks, or by only one specific task, respectively. The feature learning and the multiple task learning are iteratively optimized to form a multitask graph classification model with a global optimization goal. Experiments on real-world functional brain analysis and chemical compound categorization demonstrate the algorithm's performance. Results confirm that our method can be used to explicitly capture task correlations and uniqueness in the feature space, and explicitly answer what are shared between tasks and what is the uniqueness of a specific task.

In this paper, we advance graph classification to handle multi-graph learning for complicated objects, where each object is represented as a bag of graphs and the label is only available to each bag but not individual graphs. In addition, when training classifiers, users are only given a handful of positive bags and many unlabeled bags, and the learning objective is to train models to classify previously unseen graph bags with maximum accuracy. To achieve the goal, we propose a positive and unlabeled multi-graph learning (puMGL) framework to first select informative subgraphs to convert graphs into a feature space. To utilize unlabeled bags for learning, puMGL assigns a confidence weight to each bag and dynamically adjusts its weight value to select "reliable negative bags." A number of representative graphs, selected from positive bags and identified reliable negative graph bags, form a "margin graph pool" which serves as the base for deriving subgraph patterns, training graph classifiers, and further updating the bag weight values. A closed-loop iterative process helps discover optimal subgraphs from positive and unlabeled graph bags for learning. Experimental comparisons demonstrate the performance of puMGL for classifying real-world complicated objects.

Graphs are popularly used to represent objects with shared dependency relationships. To date, all existing graph clustering algorithms consider each node as a single attribute or a set of independent attributes, without realizing that content inside each node may also have complex structures. In this article, we formulate a new networked graph clustering task where a network contains a set of inter-connected (or networked) super-nodes, each of which is a single-attribute graph. The new super-node representation is applicable to many real-world applications, such as a citation network where each node denotes a paper whose content can be described as a graph, and citation relationships between papers form a networked graph (i.e., a super-graph). Networked graph clustering aims to find similar node groups, each of which contains nodes with similar content and structure information. The main challenge is to properly calculate the similarity between super-nodes for clustering. To solve the problem, we propose to characterize node similarity by integrating structure and content information of each super-node. To measure node content similarity, we use cosine distance by considering overlapped attributes between two super-nodes. To measure structure similarity, we propose an Attributed Random Walk Kernel (ARWK) to calculate the similarity between super-nodes. Detailed node content analysis is also included to build relationships between super-nodes with shared internal structure information, so the structure similarity can be calculated in a precise way. By integrating the structure similarity and content similarity as one matrix, the spectral clustering is used to achieve networked graph clustering. Our method enjoys sound theoretical properties, including bounded similarities and better structure similarity assessment than traditional graph clustering methods. Experiments on real-world applications demonstrate that our method significantly outperforms baseline approaches.

Graph classification is an important tool for analyzing data with structure dependency, where subgraphs are often used as features for learning. In reality, the dimension of the subgraphs crucially depends on the threshold setting of the frequency support parameter, and the number may become extremely large. As a result, subgraphs may be incrementally discovered to form a feature stream and require the underlying graph classifier to effectively discover representative subgraph features from the subgraph feature stream. In this paper, we propose a primal-dual incremental subgraph feature selection algorithm (ISF) based on a max-margin graph classifier. The ISF algorithm constructs a sequence of solutions that are both primal and dual feasible. Each primal-dual pair shrinks the dual gap and renders a better solution for the optimal subgraph feature set. To avoid bias of ISF algorithm on short-pattern subgraph features, we present a new incremental subgraph join feature selection algorithm (ISJF) by forcing graph classifiers to join short-pattern subgraphs and generate long-pattern subgraph features. We evaluate the performance of the proposed models on both synthetic networks and real-world social network data sets. Experimental results demonstrate the effectiveness of the proposed methods.

In this paper, we focus on automatically detecting events in unconstrained videos without the use of any visual training exemplars. In principle, zero-shot learning makes it possible to train an event detection model based on the assumption that events (e.g. birthday party) can be described by multiple mid-level semantic concepts (e.g. ``blowing candle'', ``birthday cake''). Towards this goal, we first pre-train a bundle of concept classifiers using data from other sources. Then we evaluate the semantic correlation of each concept w.r.t. the event of interest and pick up the relevant concept classifiers, which are applied on all test videos to get multiple prediction score vectors. While most existing systems combine the predictions of the concept classifiers with fixed weights, we propose to learn the optimal weights of the concept classifiers for each testing video by exploring a set of online available videos with free-form text descriptions of their content. To validate the effectiveness of the proposed approach, we have conducted extensive experiments on the latest TRECVID MEDTest 2014, MEDTest 2013 and CCV dataset. The experimental results confirm the superiority of the proposed approach.

Social recommendation has been widely studied in recent years. Existing social recommendation models use various explicit pieces of social information as regularization terms in recommendation, for instance, social links are considered as new constraints. However, social influence, an implicit source of information in social networks, is seldomly considered, even though it often drives recommendations in social networks. In this paper, we introduce a new global and local influence-based social recommendation model. Based on the observation that user purchase behaviour is influenced by both global influential nodes and the local influential nodes of the user, we formulate the global and local influence as an regularization terms, and incorporate them into a matrix factorization-based recommendation model. Experimental results on large data sets demonstrate the performance of the proposed method.

We address the problem of classifying sparsely labeled networks, where labeled nodes in the network are extremely scarce. Existing algorithms, such as collective classification, have been shown to be effective for jointly deriving labels of related nodes, by exploiting class label dependencies among neighboring nodes. However, when the underlying network is sparsely labeled, most nodes have too few or even no connections to labeled nodes. This makes it very difficult to leverage supervised knowledge from labeled nodes to accurately estimate label dependencies, thereby largely degrading the classification accuracy. In this paper, we propose a novel discriminative matrix factorization (DMF) based algorithm that effectively learns a latent network representation by exploiting topological paths between labeled and unlabeled nodes, in addition to nodes' content information. The main idea is to use matrix factorization to obtain a compact representation of the network that fully encodes nodes' content information and network structure, and unleash discriminative power inferred from labeled nodes to directly benefit collective classification. To achieve this, we formulate a new matrix factorization objective function that integrates network representation learning with an empirical loss minimization for classifying node labels. An efficient optimization algorithm based on conjugate gradient methods is proposed to solve the new objective function. Experimental results on real-world networks show that DMF yields superior performance gain over the state-of-the-art baselines on sparsely labeled networks.

In this paper, we present a novel transfer learning framework for network node classification. Our objective is to accurately predict node labels in a target network by leveraging information from an auxiliary source network. Such a transfer learning framework is potentially useful for broader areas of network classification, where emerging new networks might not have sufficient labeled information because node labels are either costly to obtain or simply not available, whereas many established networks from related domains are available to benefit the learning. In reality, the source and the target networks may not share common nodes or connections, so the major challenge of cross-network transfer learning is to identify knowledge/patterns transferable between networks and potentially useful to support cross-network learning. In this work, we propose to learn common signature subgraphs between networks, and use them as structure features for the target network. By combining the original node content features and the new structure features, we develop an iterative classification algorithm, TrGraph, that utilizes label dependency to jointly classify nodes in the target network. Experiments on real-world networks demonstrate that TrGraph achieves the superior performance compared to the state-of-the-art baseline methods.

We formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. More details can be found in [1].

Min-Hash, as a member of the Locality Sensitive Hashing (LSH) family for sketching sets, plays an important role in the big data era. It is widely used for efficiently estimating similarities of bag-of-words represented data and has been extended to dealing with multi-sets and real-value weighted sets. Improved Consistent Weighted Sampling (ICWS) has been recognized as the state-of-the-art for real-value weighted Min-Hash. However, the algorithmic implementation of ICWS is flawed because it violates the uniformity of the Min-Hash scheme. In this paper, we propose a Canonical Consistent Weighted Sampling (CCWS) algorithm, which not only retains the same theoretical complexity as ICWS but also strictly complies with the definition of Min-Hash. The experimental results demonstrate that the proposed CCWS algorithm runs faster than the state-of-the-arts while achieving similar classification performance on a number of real-world text data sets.

Social recommendation explores social information to improve the quality of a recommender system. It can be further divided into explicit and implicit social network recommendation. The former assumes the existence of explicit social connections between users in addition to the rating data. The latter one assumes the availability of only the ratings but not the social connections between users since the explicit social information data may not necessarily be available and usually are binary decision values (e.g., whether two people are friends), while the strength of their relationships is missing. Most of the works in this field use only rating data to infer the latent social networks. They ignore the dynamic nature of users that the preferences of users drift over time distinctly. To this end, we propose a new Implicit Dynamic Social Recommendation (IDSR) model, which infers latent social network from cascade data. It can sufficiently mine the information contained in time by mining the cascade data and identify the dynamic changes in the users in time by using the latest updated social network to make recommendations. Experiments and comparisons on three real-world datasets show that the proposed model outperforms the state-of-the-art solutions in both explicit and implicit scenarios.

Advances in social networking and communication technologies have witnessed an increasing number of applications where data is not only characterized by rich content information, but also connected with complex relationships representing social roles and dependencies between individuals. To enable knowledge discovery from such networked data, network representation learning (NRL) aims to learn vector representations for network nodes, such that off-the-shelf machine learning algorithms can be directly applied. To date, existing NRL methods either primarily focus on network structure or simply combine node content and topology for learning. We argue that in information networks, information is mainly originated from three sources: (1) homophily, (2) topology structure, and (3) node content. Homophily states social phenomenon where individuals sharing similar attributes (content) tend to be directly connected through local relational ties, while topology structure emphasizes more on global connections. To ensure effective network representation learning, we propose to augment three information sources into one learning objective function, so that the interplay roles between three parties are enforced by requiring the learned network representations (1) being consistent with node content and topology structure, and also (2) following the social homophily constraints in the learned space. Experiments on multi-class node classification demonstrate that the representations learned by the proposed method consistently outperform state-of-the-art NRL methods, especially for very sparsely labeled networks.

An enterprise social network (ESN) involves diversified user groups from producers, suppliers, logistics, to end consumers, and users have different scales, broad interests, and various objectives, such as advertising, branding, customer relationship management etc. In addition, such a highly diversified network is also featured with rich content, including recruiting messages, advertisements, news release, customer complains etc. Due to such complex nature, an immediate need is to properly organize a chaotic enterprise social network as functional groups, where each group corresponds to a set of peers with business interactions and common objectives, and further understand the business role of each group, such as their common interests and key features differing from other groups. In this paper, we argue that due to unique characteristics of enterprise social networks, simple clustering for ESN nodes or using existing topic discovery methods cannot effectively discover functional groups and understand their roles. Alternatively, we propose CENFLD, which carries out co-clustering on enterprise social networks for functional group discovery and understanding. CENFLD is a co-factorization based framework which combines network topology structures and rich content information, including interactions between nodes and correlations between node content, to discover functional user groups. Because the number of functional groups is highly data driven and hard to estimate, CENFLD employs a hold-out test principle to find the group number optimally complying with the underlying data. Experiments and comparisons, with state-of-the-art approaches, on 13 real-world enterprise/organizational networks validate the performance of CENFLD.

In the research of building emotion lexicons, we witness the exploitation of crowd-sourced affective annotation given by readers of online news articles. Such approach ignores the relationship between topics and emotion expressions which are often closely correlated. We build an emotion lexicon by developing a novel joint non-negative matrix factorization model which not only incorporates crowd-annotated emotion labels of articles but also generates the lexicon using the topic-specific matrices obtained from the factorization process. We evaluate our lexicon via emotion classification on both benchmark and built-in-house datasets. Results demonstrate the high-quality of our lexicon.

Principal component analysis (PCA) has been widely applied to dimensionality reduction and data pre-processing for different applications in engineering, biology, social science, and the like. Classical PCA and its variants seek for linear projections of the original variables to obtain the low-dimensional feature representations with maximal variance. One limitation is that it is difficult to interpret the results of PCA. Besides, the classical PCA is vulnerable to certain noisy data. In this paper, we propose a Convex Sparse Principal Component Analysis (CSPCA) algorithm and apply it to feature learning. First, we show that PCA can be formulated as a low-rank regression optimization problem. Based on the discussion, thel2, 1-normminimization is incorporated into the objective function to make the regression coefficients sparse, thereby robust to the outliers. Also, based on the sparse model used in CSPCA, an optimal weight is assigned to each of the original feature, which in turn provides the output with good interpretability. With the output of our CSPCA, we can effectively analyze the importance of each feature under the PCA criteria. Our new objective function is convex, and we propose an iterative algorithm to optimize it. We apply the CSPCA algorithm to feature selection and conduct extensive experiments on seven benchmark datasets. Experimental results demonstrate that the proposed algorithm outperforms state-of-the-art unsupervised feature selection algorithms.

Graph classification aims to learn models to classify structure data. To date, all existing graph classification methods are designed to target one single learning task and require a large number of labeled samples for learning good classification models. In reality, each real-world task may only have a limited number of labeled samples, yet multiple similar learning tasks can provide useful knowledge to benefit all tasks as a whole. In this paper, we formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. The niche of MTG stems from the fact that with a limited number of training samples, subgraph features selected for one single graph classification task tend to overfit the training data. By using additional tasks as evaluation sets, MTG can jointly regularize multiple tasks to explore high quality subgraph features for graph classification. To achieve this goal, we formulate an objective function which combines multiple graph classification tasks to evaluate the informativeness score of a subgraph feature. An iterative subgraph feature exploration and multi-task learning process is further proposed to incrementally select subgraph features for graph classification. Experiments on real-world multi-task graph classification datasets demonstrate significant performance gain.

In this paper, we study a new problem of continuous learning from doubly-streaming data where both data volume and feature space increase over time. We refer to the doubly-streaming data as trapezoidal data streams and the corresponding learning problem as online learning from trapezoidal data streams. The problem is challenging because both data volume and data dimension increase over time, and existing online learning[1],[2], online feature selection[3], and streaming feature selection algorithms[4],[5]are inapplicable. We propose a new Online Learning with Streaming Features algorithm (OL$_{SF}$for short) and its two variants, which combine online learning[1],[2]and streaming feature selection[4],[5]to enable learning from trapezoidal data streams with infinite training instances and features. When a new training instance carrying new features arrives, a classifier updates the existing features by following the passive-aggressive update rule[2]and updates the new features by following the structural risk minimization principle. Feature sparsity is then introduced by using the projected truncation technique. We derive performance bounds of the OL$_{SF}$algorithm and its variants. We also conduct experiments on real-world data sets to show the performance of the proposed algorithms.

In many real-world applications, data are represented by matrices or high-order tensors. Despite the promising performance, the existing 2-D discriminant analysis algorithms employ a single projection model to exploit the discriminant information for projection, making the model less flexible. In this paper, we propose a novel compound rank- k projection (CRP) algorithm for bilinear analysis. The CRP deals with matrices directly without transforming them into vectors, and it, therefore, preserves the correlations within the matrix and decreases the computation complexity. Different from the existing 2-D discriminant analysis algorithms, objective function values of CRP increase monotonically. In addition, the CRP utilizes multiple rank- k projection models to enable a larger search space in which the optimal solution can be found. In this way, the discriminant ability is enhanced. We have tested our approach on five data sets, including UUIm, CVL, Pointing'04, USPS, and Coil20. Experimental results show that the performance of our proposed CRP performs better than other algorithms in terms of classification accuracy.

Witnessing the wide spread of malicious information in large networks, we develop an efficient method to detect anomalous diffusion sources and thus protect networks from security and privacy attacks. To date, most existing work on diffusion sources detection are based on the assumption that network snapshots that reflect information diffusion can be obtained continuously. However, obtaining snapshots of an entire network needs to deploy detectors on all network nodes and thus is very expensive. Alternatively, in this article, we study the diffusion sources locating problem by learning from information diffusion data collected from only a small subset of network nodes. Specifically, we present a new regression learning model that can detect anomalous diffusion sources by jointly solving five challenges, that is, unknown number of source nodes, few activated detectors, unknown initial propagation time, uncertain propagation path and uncertain propagation time delay. We theoretically analyze the strength of the model and derive performance bounds. We empirically test and compare the model using both synthetic and real-world networks to demonstrate its performance.

Graph classification is an important tool for analysing structured and semi-structured data, where subgraphs are commonly used as the feature representation. However, the number and size of subgraph features crucially depend on the threshold parameters of frequent subgraph mining algorithms. Any improper setting of the parameters will generate many trivial short-pattern subgraph fragments which dominate the feature space, distort graph classifiers and bury interesting long-pattern subgraphs. In this paper, we propose a new Subgraph Join Feature Selection (SJFS) algorithm. The SJFS algorithm, by forcing graph classifiers to join short-pattern subgraph fragments, can defrag trivial subgraph features and deliver long-pattern interesting subgraphs. Experimental results on both synthetic and real-world social network graph data demonstrate the performance of the proposed method.

We study a new problem of learning from doubly-streaming data where both data volume and feature space increase over time. We refer to the problem as mining trapezoidal data streams. The problem is challenging because both data volume and feature space are increasing, to which existing online learning, online feature selection and streaming feature selection algorithms are inapplicable. We propose a new Sparse Trapezoidal Streaming Data mining algorithm (STSD) and its two variants which combine online learning and online feature selection to enable learning trapezoidal data streams with infinite training instances and features. Specifically, when new training instances carrying new features arrive, the classifier updates the existing features by following the passive-aggressive update rule used in online learning and updates the new features with the structural risk minimization principle. Feature sparsity is also introduced using the projected truncation techniques. Extensive experiments on the demonstrated UCI data sets show the performance of the proposed algorithms.

In this paper we study a new problem of online diffusion source detection in social networks. Existing work on diffusion source detection focuses on offline learning, which assumes data collected from network detectors are static and a snapshot of network is available before learning. However, an offline learning model does not meet the needs of early warning, real-time awareness, and real-time response of malicious information spreading in social networks. In this paper, we combine online learning and regression-based detection methods for real-time diffusion source detection. Specifically, we propose a new ℓ1non-convex regression model as the learning function, and an Online Stochastic Sub-gradient algorithm (OSS for short). The proposed model is empirically evaluated on both synthetic and real-world networks. Experimental results demonstrate the effectiveness of the proposed model.

Mining dense subgraphs from a large graph is a fundamental graph mining task and can be widely applied in a variety of application domains such as network science, biology, graph database, web mining, graph compression, and micro-blogging systems. Here a dense subgraph is defined as a subgraph with high density (#.edge / #.node). Existing studies of this problem either focus on finding the densest subgraph or identifying an optimal clique-like dense subgraph, and they adopt a simple greedy approach to find the top-k dense subgraphs. However, their identified subgraphs cannot be used to represent the dense regions of the graph. Intuitively, to represent a dense region, the subgraph identified should be the subgraph with highest density in its local region in the graph. However, it is non-trivial to formally model a locally densest subgraph. In this paper, we aim to discover top-k such representative locally densest subgraphs of a graph. We provide an elegant parameter-free definition of a locally densest subgraph. The definition not only fits well with the intuition, but is also associated with several nice structural properties. We show that the set of locally densest subgraphs in a graph can be computed in polynomial time. We further propose three novel pruning strategies to largely reduce the search space of the algorithm. In our experiments, we use several real datasets with various graph properties to evaluate the effectiveness of our model using four quality measures and a case study. We also test our algorithms on several real web-scale graphs, one of which contains 118.14 million nodes and 1.02 billion edges, to demonstrate the high efficiency of the proposed algorithms.

Cross-domain collaborative filtering (CF) aims to share common rating knowledge across multiple related CF domains to boost the CF performance. In this paper, we view CF domains as a 2-D site-time coordinate system, on which multiple related domains, such as similar recommender sites or successive time-slices, can share group-level rating patterns. We propose a unified framework for cross-domain CF over the site-time coordinate system by sharing group-level rating patterns and imposing user/item dependence across domains. A generative model, say ratings over site-time (ROST), which can generate and predict ratings for multiple related CF domains, is developed as the basic model for the framework. We further introduce cross-domain user/item dependence into ROST and extend it to two real-world cross-domain CF scenarios: 1) ROST (sites) for alleviating rating sparsity in the target domain, where multiple similar sites are viewed as related CF domains and some items in the target domain depend on their correspondences in the related ones; and 2) ROST (time) for modeling user-interest drift over time, where a series of time-slices are viewed as related CF domains and a user at current time-slice depends on herself in the previous time-slice. All these ROST models are instances of the proposed unified framework. The experimental results show that ROST (sites) can effectively alleviate the sparsity problem to improve rating prediction performance and ROST (time) can clearly track and visualize user-interest drift over time.

Many applications involve stream data with structural dependency, graph representations, and continuously increasing volumes. For these applications, it is very common that their class distributions are imbalanced with minority (or positive) samples being only a small portion of the population, which imposes significant challenges for learning models to accurately identify minority samples. This problem is further complicated with the presence of noise, because they are similar to minority samples and any treatment for the class imbalance may falsely focus on the noise and result in deterioration of accuracy. In this paper, we propose a classification model to tackle imbalanced graph streams with noise. Our method, graph ensemble boosting, employs an ensemble-based framework to partition graph stream into chunks each containing a number of noisy graphs with imbalanced class distributions. For each individual chunk, we propose a boosting algorithm to combine discriminative subgraph pattern selection and model learning as a unified framework for graph classification. To tackle concept drifting in graph streams, an instance level weighting mechanism is used to dynamically adjust the instance weight, through which the boosting framework can emphasize on difficult graph samples. The classifiers built from different graph chunks form an ensemble for graph stream classification. Experiments on real-life imbalanced graph streams demonstrate clear benefits of our boosting design for handling imbalanced noisy graph stream.

This article proposes LA-LDA, a location-aware probabilistic generative model that exploits location-based ratings to model user profiles and produce recommendations. Most of the existing recommendation models do not consider the spatial information of users or items; however, LA-LDA supports three classes of location-based ratings, namely spatial user ratings for nonspatial items, nonspatial user ratings for spatial items, and spatial user ratings for spatial items. LA-LDA consists of two components, ULA-LDA and ILA-LDA, which are designed to take into account user and item location information, respectively. The component ULA-LDA explicitly incorporates and quantifies the influence from local public preferences to produce recommendations by considering user home locations, whereas the component ILA-LDA recommends items that are closer in both taste and travel distance to the querying users by capturing item co-occurrence patterns, as well as item location co-occurrence patterns. The two components of LA-LDA can be applied either separately or collectively, depending on the available types of location-based ratings. To demonstrate the applicability and flexibility of the LA-LDA model, we deploy it to both top- k recommendation and cold start recommendation scenarios. Experimental evidence on large-scale real-world data, including the data from Gowalla (a location-based social network), DoubanEvent (an event-based social network), and MovieLens (a movie recommendation system), reveal that LA-LDA models user profiles more accurately by outperforming existing recommendation models for top- k recommendation and the cold start problem.

In this paper, we present a novel transfer learning framework for network node classification. Our objective is to accurately predict the labels of nodes in a target network by leveraging information from an auxiliary source network. Such a transfer learning framework is potentially useful for broader areas of network classification, where emerging new networks might not have sufficient labeled information because node labels are either costly to obtain or simply not available, whereas many established networks from related domains are available to benefit the learning. In reality, the source and the target networks may not share common nodes or connections, so the major challenge of cross-network transfer learning is to identify knowledge/patterns transferable between networks and potentially useful to support cross-network learning. In this work, we propose to learn common signature subgraphs between networks, and use them to construct new structure features for the target network. By combining the original node content features and the new structure features, we develop an iterative classification algorithm, TrGraph, that utilizes label dependency to jointly classify nodes in the target network. Experiments on real-world networks demonstrate that TrGraph achieves the superior performance compared to the state-of-the-art baseline methods, and transferring generalizable structure information can indeed improve the node classification accuracy.

In traditional multi-instance learning (MIL), instances are typically represented by using a single feature view. As MIL becoming popular in domain specific learning tasks, aggregating multiple feature views to represent multi-instance bags has recently shown promising results, mainly because multiple views provide extra information for MIL tasks. Nevertheless, multiple views also increase the risk of involving redundant views and irrelevant features for learning. In this paper, we formulate a new cross-view feature selection problem that aims to identify the most representative features across all feature views for MIL. To achieve the goal, we design a new optimization problem by integrating both multi-view representation and multi-instance bag constraints. The solution to the objective function will ensure that the identified top-m features are the most informative ones across all feature views. Experiments on two real-world applications demonstrate the performance of the cross-view feature selection for content-based image retrieval and social media content recommendation.

Many real-world networks are featured with dynamic changes, such as new nodes and edges, and modification of the node content. Because changes are continuously introduced to the network in a streaming fashion, we refer to such dynamic networks as streaming networks. In this paper, we propose a new classification method for streaming networks, namely streaming network node classification (SNOC). For streaming networks, the essential challenge is to properly capture the dynamic changes of the node content and node interactions to support node classification. While streaming networks are dynamically evolving, for a short temporal period, a subset of salient features are essentially tied to the network content and structures, and therefore can be used to characterize the network for classification. To achieve this goal, we propose to carry out streaming network feature selection (SNF) from the network, and use selected features as gauge to classify unlabeled nodes. A Laplacian based quality criterion is proposed to guide the node classification, where the Laplacian matrix is generated based on node labels and structures. Node classification is achieved by finding the class that results in the minimal gauging value with respect to the selected features. By frequently updating the features selected from the network, node classification can quickly adapt to the changes in the network for maximal performance gain. Experiments demonstrate that SNOC is able to capture changes in network structures and node content, and outperforms baseline approaches with significant performance gain.

Graph classification has traditionally focused on graphs generated from a single feature view. In many applications, it is common to have useful information from different channels/views to describe objects, which naturally results in a new representation with multiple graphs generated from different feature views being used to describe one object. In this paper, we formulate a new Multi-Graph-View learning task for graph classification, where each object to be classified contains graphs from multiple graph-views. This problem setting is essentially different from traditional single-graph-view graph classification, where graphs are from one single feature view. To solve the problem, we propose a Cross Graph-View Sub graph Feature based Learning (gCGVFL) algorithm that explores an optimal set of sub graphs, across multiple graph-views, as features to represent graphs. Specifically, we derive an evaluation criterion to estimate the discriminative power and the redundancy of sub graph features across all views, and assign proper weight values to each view to indicate its importance for graph classification. The iterative cross graph-view sub graph scoring and graph-view weight updating form a closed loop to find optimal sub graphs to represent graphs for multi-graph-view learning. Experiments and comparisons on real-world tasks demonstrate the algorithm's performance.

A Bayesian Network (BN) is a graphical model which can be used to represent conditional dependency between random variables, such as diseases and symptoms. A Bayesian Network Classifier (BNC) uses BN to characterize the relationships between attributes and the class labels, where a simplified approach is to employ a conditional independence assumption between attributes and the corresponding class labels, i.e., the Naive Bayes (NB) classification model. One major approach to mitigate NB's primary weakness (the conditional independence assumption) is the attribute weighting, and this type of approach has been proved to be effective for NB with simple structure. However, for weighted BNCs involving complex structures, in which attribute weighting is embedded into the model, there is no existing study on whether the weighting will work for complex BNCs and how effective it will impact on the learning of a given task. In this paper, we first survey several complex structure models for BNCs, and then carry out experimental studies to investigate the effectiveness of the attribute weighting strategies for complex BNCs, with a focus on Hidden Naive Bayes (HNB) and Averaged One-Dependence Estimation (AODE). Our studies use classification accuracy (ACC), area under the ROC curve ranking (AUC), and conditional log likelihood (CLL), as the performance metrics. Experiments and comparisons on 36 benchmark data sets demonstrate that attribute weighting technologies just slightly outperforms unweighted complex BNCs with respect to the ACC and AUC, but significant improvement can be observed using CLL.

Naive Bayes (NB) network is a popular classification technique for data mining and machine learning. Many methods exist to improve the performance of NB by overcoming its primary weakness - the assumption that attributes are conditionally independent given the class, using techniques such as backwards sequential elimination and lazy elimination. Some weighting technologies, including attribute weighting and instance weighting, have also been proposed to improve the accuracy of NB. In this paper, we propose a dual weighted model, namely DWNB, for NB classification. In DWNB, we firstly employ an instance similarity based method to weight each training instance. After that, we build an attribute weighted model based on the new training data, where the calculation of the probability value is based on the embedded instance weights. The dual instance and attribute weighting allows DWNB to tackle the conditional independence assumption for accurate classification. Experiments and comparisons on 36 benchmark data sets demonstrate that DWNB outperforms existing weighted NB algorithms.

In this paper, we formulate a new multi-graph learning task with only positive and unlabeled bags, where labels are only available for bags but not for individual graphs inside the bag. This problem setting raises significant challenges because bag-of-graph setting does not have features to directly represent graph data, and no negative bags exits for deriving discriminative classification models. To solve the challenge, we propose a puMGL learning framework which relies on two iteratively combined processes for multigraph learning: (1) deriving features to represent graphs for learning; and (2) deriving discriminative models with only positive and unlabeled graph bags. For the former, we derive a subgraph scoring criterion to select a set of informative subgraphs to convert each graph into a feature space. To handle unlabeled bags, we assign a weight value to each bag and use the adjusted weight values to select most promising unlabeled bags as negative bags. A margin graph pool (MGP), which contains some representative graphs from positive bags and identified negative bags, is used for selecting subgraphs and training graph classifiers. The iterative subgraph scoring, bag weight updating, and MGP based graph classification forms a closed loop to find optimal subgraphs and most suitable unlabeled bags for multi-graph learning. Experiments and comparisons on real-world multigraph data demonstrate the algorithm performance.

MapReduce has become one of the most popular parallel computing paradigms in cloud, due to its high scalability, reliability, and fault-tolerance achieved for a large variety of applications in big data processing. In the literature, there are MapReduce Class MRC and Minimal MapReduce Class MMC to define the memory consumption, communication cost, CPU cost, and number of MapReduce rounds for an algorithm to execute in MapReduce. However, neither of them is designed for big graph processing in MapReduce, since the constraints in MMC can be hardly achieved simultaneously on graphs and the conditions in MRC may induce scalability problems when processing big graph data. In this paper, we study scalable big graph processing in MapReduce. We introduce a Scalable Graph processing Class SGC by relaxing some constraints in MMC to make it suitable for scalable graph processing. We define two graph join operators in SGC, namely, EN join and NE join, using which a wide range of graph algorithms can be designed, including PageRank, breadth first search, graph keyword search, Connected Component (CC) computation, and Minimum Spanning Forest (MSF) computation. Remarkably, to the best of our knowledge, for the two fundamental graph problems CC and MSF computation, this is the first work that can achieve O(log(n)) MapReduce rounds with $O(n+m)$ total communication cost in each round and constant memory consumption on each machine, where $n$ and $m$ are the number of nodes and edges in the graph respectively. We conducted extensive performance studies using two web-scale graphs Twitter and Friendster with different graph characteristics. The experimental results demonstrate that our algorithms can achieve high scalability in big graph processing.

The dominance operator plays an important role in a wide spectrum of multi-criteria decision making applications. Generally speaking, a dominance operator is a partial order on a set O of objects, and we say the dominance operator has the monotonic property regarding a family of ranking functions F if o1 dominates o2 implies f(o1) ≥ f(o2) for any ranking function f ∈ F and objects o1, o2 ∈ O. The dominance operator on the multi-dimensional points is well defined, which has the monotonic property regarding any monotonic ranking (scoring) function. Due to the uncertain nature of data in many emerging applications, a variety of existing works have studied the semantics of ranking query on uncertain objects. However, the problem of dominance operator against multi-dimensional uncertain objects remains open. Although there are several attempts to propose dominance operator on multi-dimensional uncertain objects, none of them claims the monotonic property on these ranking approaches.

Traditional active learning methods require the labeler to provide a class label for each queried instance. The labelers are normally highly skilled domain experts to ensure the correctness of the provided labels, which in turn results in expensive labeling cost. To reduce labeling cost, an alternative solution is to allow nonexpert labelers to carry out the labeling task without explicitly telling the class label of each queried instance. In this paper, we propose a new active learning paradigm, in which a nonexpert labeler is only asked “whether a pair of instances belong to the same class”, namely, a pairwise label homogeneity. Under such circumstances, our active learning goal is twofold: (1) decide which pair of instances should be selected for query, and (2) how to make use of the pairwise homogeneity information to improve the active learner. To achieve the goal, we propose a “Pairwise Query on Max-flow Paths” strategy to query pairwise label homogeneity from a nonexpert labeler, whose query results are further used to dynamically update a Min-cut model (to differentiate instances in different classes). In addition, a “Confidence-based Data Selection” measure is used to evaluate data utility based on the Min-cut model’s prediction results. The selected instances, with inferred class labels, are included into the labeled set to form a closed-loop active learning process. Experimental results and comparisons with state-of-the-art methods demonstrate that our new active learning paradigm can result in good performance with nonexpert labelers.

This paper formulates a multi-graph learning task. In our problem setting, a bag contains a number of graphs and a class label. A bag is labeled positive if at least one graph in the bag is positive, and negative otherwise. In addition, the genuine label of each graph in a positive bag is unknown, and all graphs in a negative bag are negative. The aim of multi-graph learning is to build a learning model from a number of labeled training bags to predict previously unseen test bags with maximum accuracy. This problem setting is essentially different from existing multi-instance learning (MIL), where instances in MIL share well-defined feature values, but no features are available to represent graphs in a multi-graph bag. To solve the problem, we propose a Multi-Graph Feature based Learning (gMGFL) algorithm that explores and selects a set of discriminative subgraphs as features to transfer each bag into a single instance, with the bag label being propagated to the transferred instance. As a result, the multi-graph bags form a labeled training instance set, so generic learning algorithms, such as decision trees, can be used to derive learning models for multi-graph classification. Experiments and comparisons on real-world multi-graph tasks demonstrate the algorithm performance.

Data uncertainty has posed many unique challenges to nearly all types of data mining tasks, creating a need for uncertain data mining. In this paper, we focus on the particular task of mining probabilistic frequent serial episodes (P-FSEs) from uncertain sequence data, which applies to many real applications including sensor readings as well as customer purchase sequences. We first define the notion of P-FSEs, based on the frequentness probabilities of serial episodes under possible world semantics. To discover P-FSEs over an uncertain sequence, we propose: 1) an exact approach that computes the accurate frequentness probabilities of episodes; 2) an approximate approach that approximates the frequency of episodes using probability models; 3) an optimized approach that efficiently prunes a candidate episode by estimating an upper bound of its frequentness probability using approximation techniques.

Graph classification is becoming increasingly popular due to the rapidly rising applications involving data with structural dependency. The wide spread of the graph applications and the inherent complex relationships between graph objects have made the labels of the graph data expensive and/or difficult to obtain, especially for applications involving dynamic changing graph records. While labeled graphs are limited, the copious amounts of unlabeled graphs are often easy to obtain with trivial efforts. In this paper, we propose a framework to build a stream based graph classification model by combining both labeled and unlabeled graphs. Our method, called gSLU, employs an ensemble based framework to partition graph streams into a number of graph chunks each containing some labeled and unlabeled graphs. For each individual chunk, we propose a minimum-redundancy subgraph feature selection module to select a set of informative subgraph features to build a classifier. To tackle the concept drifting in graph streams, an instance level weighting mechanism is used to dynamically adjust the instance weight, through which the subgraph feature selection can emphasize on difficult graph samples. The classifiers built from different graph chunks form an ensemble for graph stream classification. Experiments on real-world graph streams demonstrate clear benefits of using minimum-redundancy subgraph features to build accurate classifiers. By employing instance level weighting, our graph ensemble model can effectively adapt to the concept drifting in the graph stream for classification.

In this paper, we focus on the problem of mining Probabilistic Dependent Frequent Serial Episodes (P-DFSEs) from uncertain sequence data. By observing that the frequentness probability of an episode in an uncertain sequence is a Markov Chain imbeddable variable, we first propose an Embeded Markov Chain-based algorithm that efficiently computes the frequentness probability of an episode by projecting the probability space into a set of limited partitions. To further improve the computation efficiency, we devise an optimized approach that prunes candidate episodes early by estimating the upper bound of their frequentness probabilities.

Multi-instance learning concerns about building learning models from a number of labeled instance bags, where each bag consists of instances with unknown labels. A bag is labeled positive if one or more multiple instances inside the bag is positive, and negative otherwise. For all existing multi-instance learning algorithms, they are only applicable to the setting where instances in each bag are represented by a set of well defined feature values. In this paper, we advance the problem to a multi-instance multi-graph setting, where a bag contains a number of instances and graphs in pairs, and the learning objective is to derive classification models from labeled bags, containing both instances and graphs, to predict previously unseen bags with maximum accuracy. To achieve the goal, the main challenge is to properly represent graphs inside each bag and further take advantage of complementary information between instance and graph pairs for learning. In the paper, we propose a Dual Embedding Multi-Instance Multi-Graph Learning (DE-MIMG) algorithm, which employs a dual embedding learning approach to (1) embed instance distributions into the informative sub graphs discovery process, and (2) embed discovered sub graphs into the instance feature selection process. The dual embedding process results in an optimal representation for each bag to provide combined instance and graph information for learning. Experiments and comparisons on real-world multi-instance multi-graph learning tasks demonstrate the algorithm performance.

Mining probabilistic frequent patterns from uncertain data has received a great deal of attention in recent years due to the wide applications. However, probabilistic frequent pattern mining suffers from the problem that an exponential number of result patterns are generated, which seriously hinders further evaluation and analysis. In this paper, we focus on the problem of mining probabilistic representative frequent patterns (P-RFP), which is the minimal set of patterns with adequately high probability to represent all frequent patterns. Observing the bottleneck in checking whether a pattern can probabilistically represent another, which involves the computation of a joint probability of the supports of two patterns, we introduce a novel approximation of the joint probability with both theoretical and empirical proofs. Based on the approximation, we propose an Approximate P-RFP Mining (APM) algorithm, which effectively and efficiently compresses the set of probabilistic frequent patterns. To our knowledge, this is the first attempt to analyze the relationship between two probabilistic frequent patterns through an approximate approach. Our experiments on both synthetic and real-world datasets demonstrate that the APM algorithm accelerates P-RFP mining dramatically, orders of magnitudes faster than an exact solution. Moreover, the error rate of APM is guaranteed to be very small when the database contains hundreds transactions, which further affirms APM is a practical solution for summarizing probabilistic frequent patterns.

Probabilistic frequent pattern mining over uncertain data has received a great deal of attention recently due to the wide applications of uncertain data. Similar to its counterpart in deterministic databases, however, probabilistic frequent pattern mining suffers from the same problem of generating an exponential number of result patterns. The large number of discovered patterns hinders further evaluation and analysis, and calls for the need to find a small number of representative patterns to approximate all other patterns. This paper formally defines the problem of probabilistic representative frequent pattern (P-RFP) mining, which aims to find the minimal set of patterns with sufficiently high probability to represent all other patterns. The problem's bottleneck turns out to be checking whether a pattern can probabilistically represent another, which involves the computation of a joint probability of supports of two patterns. To address the problem, we propose a novel and efficient dynamic programming-based approach. Moreover, we have devised a set of effective optimization strategies to further improve the computation efficiency. Our experimental results demonstrate that the proposed P-RFP mining effectively reduces the size of probabilistic frequent patterns. Our proposed approach not only discovers the set of P-RFPs efficiently, but also restores the frequency probability information of patterns with an error guarantee.

With the recent explosion of social network applications, active learning has increasingly become an important paradigm for classifying networked data. While existing research has shown promising results by exploiting network properties to improve the active learning performance, they are all based on a static setting where the number and the type of classes underlying the networked data remain stable and unchanged. For most social network applications, the dynamic change of users and their evolving relationships, along with the emergence of new social events, often result in new classes that need to be immediately discovered and labeled for classification. This paper proposes a novel approach called ADLNET for active class discovery and learning with networked data. Our proposed method uses the Dirichlet process defined over class distributions to enable active discovery of new classes, and explicitly models label correlations in the utility function of active learning. Experimental results on two real-world networked data sets demonstrate that our proposed approach outperforms other state-of-the-art methods.

Active learning traditionally assumes that an oracle is capable of providing labeling information for each query instance. This paper formulates a new research problem which allows an oracle admit that he/she is incapable of labeling some query instances or simply answer "I don't know the label." We define a unified objectivefunction to ensure that each query instance submitted to the oracleis the one mostly needed for labeling and the oracle should also hasthe knowledge to label. Experiments based on different types of knowledge blind spot (KBS) models demonstrate the effectiveness of theproposed design.

Imbalanced time series classification (TSC) involving many real-world applications has increasingly captured attention of researchers. Previous work has proposed an intelligent-structure preserving over-sampling method (SPO), which the authors claimed achieved better performance than other existing over-sampling and state-of-the-art methods in TSC. The main disadvantage of over-sampling methods is that they significantly increase the computational cost of training a classification model due to the addition of new minority class instances to balance data-sets with high dimensional features. These challenging issues have motivated us to find a simple and efficient solution for imbalanced TSC. Statistical tests are applied to validate our conclusions. The experimental results demonstrate that this proposed simple random under-sampling technique with SVM is efficient and can achieve results that compare favorably with the existing complicated SPO method for imbalanced TSC.

Short & sparse text is becoming more prevalent on the web, such as search snippets, micro-blogs and product reviews. Accurately classifying short & sparse text has emerged as an important while challenging task. Existing work has considered utilizing external data (e.g. Wikipedia) to alleviate data sparseness, by appending topics detected from external data as new features. However, training a classifier on features concatenated from different spaces is not easy considering the features have different physical meanings and different significance to the classification task. Moreover, it exacerbates the "curse of dimensionality" problem. In this study, we propose a transfer classification method, TCSST, to exploit the external data to tackle the data sparsity issue. The transfer classifier will be learned in the original feature space. Considering that the labels of the external data may not be readily available or sufficiently enough, TCSST further exploits the unlabeled external data to aid the transfer classification. We develop novel strategies to allow TCSST to iteratively select high quality unlabeled external data to help with the classification. We evaluate the performance of TCSST on both benchmark as well as real-world data sets. Our experimental results demonstrate that the proposed method is effective in classifying very short & sparse text, consistently outperforming existing and baseline methods.

Most studies on graph classification focus on designing fast and effective kernels. Several fast subtree kernels have achieved a linear time-complexity w.r.t. the number of edges under the condition that a common feature space (e.g., a subtree pattern list) is needed to represent all graphs. This will be infeasible when graphs are presented in a stream with rapidly emerging subtree patterns. In this case, computing a kernel matrix for graphs over the entire stream is difficult since the graphs in the expired chunks cannot be projected onto the unlimitedly expanding feature space again. This leads to a big trouble for graph classification over streams -- Different portions of graphs have different feature spaces. In this paper, we aim to enable large-scale graph classification over streams using the classical ensemble learning framework, which requires the data in different chunks to be in the same feature space. To this end, we propose a Nested Subtree Hashing (NSH) algorithm to recursively project the multi-resolution subtree patterns of different chunks onto a set of common low-dimensional feature spaces. We theoretically analyze the derived NSH kernel and obtain a number of favorable properties: 1) The NSH kernel is an unbiased and highly concentrated estimator of the fast subtree kernel. 2) The bound of convergence rate tends to be tighter as the NSH algorithm steps into a higher resolution. 3) The NSH kernel is robust in tolerating concept drift between chunks over a stream. We also empirically test the NSH kernel on both a large-scale synthetic graph data set and a real-world chemical compounds data set for anticancer activity prediction. The experimental results validate that the NSH kernel is indeed efficient and robust for graph classification over streams.

The luminance of a natural scene is often of high dynamic range (HDR). In this paper, we propose a new scheme to handle HDR scenes by integrating locally adaptive scene detail capture and suppressing gradient reversals introduced by the local adaptation. The proposed scheme is novel for capturing an HDR scene by using a standard dynamic range (SDR) device and synthesizing an image suitable for SDR displays. In particular, we use an SDR capture device to record scene details (i.e., the visible contrasts and the scene gradients) in a series of SDR images with different exposure levels. Each SDR image responds to a fraction of the HDR and partially records scene details. With the captured SDR image series, we first calculate the image luminance levels, which maximize the visible contrasts, and then the scene gradients embedded in these images. Next, we synthesize an SDR image by using a probabilistic model that preserves the calculated image luminance levels and suppresses reversals in the image luminance gradients. The synthesized SDR image contains much more scene details than any of the captured SDR image. Moreover, the proposed scheme also functions as the tone mapping of an HDR image to the SDR image, and it is superior to both global and local tone mapping operators. This is because global operators fail to preserve visual details when the contrast ratio of a scene is large, whereas local operators often produce halos in the synthesized SDR image. The proposed scheme does not require any human interaction or parameter tuning for different scenes. Subjective evaluations have shown that it is preferred over a number of existing approaches.

Bagging is a simple yet effective design which combines multiple single learners to form an ensemble for prediction. Despite its popular usage in many real-world applications, existing research is mainly concerned with studying unstable learners as the key to ensure the performance gain of a bagging predictor, with many key factors remaining unclear. For example, it is not clear when a bagging predictor can outperform a single learner and what is the expected performance gain when different learning algorithms were used to form a bagging predictor. In this paper, we carry out comprehensive empirical studies to evaluate bagging predictors by using 12 different learning algorithms and 48 benchmark data-sets. Our analysis uses robustness and stability decompositions to characterize different learning algorithms, through which we rank all learning algorithms and comparatively study their bagging predictors to draw conclusions. Our studies assert that both stability and robustness are key requirements to ensure the high performance for building a bagging predictor. In addition, our studies demonstrated that bagging is statistically superior to most single base learners, except for KNN and Naïve Bayes (NB). Multi-layer perception (MLP), Naïve Bayes Trees (NBTree), and PART are the learning algorithms with the best bagging performance.

Mining Negative Sequential Patterns (NSP) is much more challenging than mining Positive Sequential Patterns (PSP) due to the high computational complexity and huge search space required in calculating Negative Sequential Candidates (NSC). Very few approaches are available for mining NSP, which mainly rely on re-scanning databases after identifying PSP. As a result, they are very inefficient. In this paper, we propose an efficient algorithm for mining NSP, called e-NSP, which mines for NSP by only involving the identified PSP, without re-scanning databases. First, negative containment is defined to determine whether or not a data sequence contains a negative sequence. Second, an efficient approach is proposed to convert the negative containment problem to a positive containment problem. The supports of NSC are then calculated based only on the corresponding PSP. Finally, a simple but efficient approach is proposed to generate NSC. With e-NSP, mining NSP does not require additional database scans, and the existing PSP mining algorithms can be integrated into e-NSP to mine for NSP efficiently. e-NSP is compared with two currently available NSP mining algorithms on 14 synthetic and real-life datasets. Intensive experiments show that e-NSP takes as little as 3% of the runtime of the baseline approaches and is applicable for efficient mining of NSP in large datasets.

Traditional active learning methods request experts to provide ground truths to the queried instances, which can be expensive in practice. An alternative solution is to ask nonexpert labelers to do such labeling work, which can not tell the definite class labels. In this paper, we propose a new active learning paradigm, in which a nonexpert labeler is only asked "whether a pair of instances belong to the same class". To instantiate the proposed paradigm, we adopt the MinCut algorithm as the base classifier. We first construct a graph based on the pairwise distance of all the labeled and unlabeled instances and then repeatedly update the unlabeled edge weights on the max-flow paths in the graph. Finally, we select an unlabeled subset of nodes with the highest prediction confidence as the labeled data, which are included into the labeled data set to learn a new classifier for the next round of active learning. The experimental results and comparisons, with state-of-the-art methods, demonstrate that our active learning paradigm can result in good performance with nonexpert labelers.

Collaborative filtering (CF) techniques recommend items to users based on their historical ratings. In real-world scenarios, user interests may drift over time since they are affected by moods, contexts, and pop culture trends. This leads to the fact that a user's historical ratings comprise many aspects of user interests spanning a long time period. However, at a certain time slice, one user's interest may only focus on one or a couple of aspects. Thus, CF techniques based on the entire historical ratings may recommend inappropriate items. In this paper, we consider modeling user-interest drift over time based on the assumption that each user has multiple counterparts over temporal domains and successive counterparts are closely related. We adopt the cross-domain CF framework to share the static group-level rating matrix across temporal domains, and let user-interest distribution over item groups drift slightly between successive temporal domains. The derived method is based on a Bayesian latent factor model which can be inferred using Gibbs sampling. Our experimental results show that our method can achieve state-of-the-art recommendation performance as well as explicitly track and visualize user-interest drift over time.

Positive and unlabelled learning (PU learning) has been investigated to deal with the situation where only the positive examples and the unlabelled examples are available. Most of the previous works focus on identifying some negative examples from the unlabelled data, so that the supervised learning methods can be applied to build a classifier. However, for the remaining unlabelled data, which can not be explicitly identified as positive or negative (we call them ambiguous examples), they either exclude them from the training phase or simply enforce them to either class. Consequently, their performance may be constrained. This paper proposes a novel approach, called similarity-based PU learning (SPUL) method, by associating the ambiguous examples with two similarity weights, which indicate the similarity of an ambiguous example towards the positive class and the negative class, respectively. The local similarity-based and global similarity-based mechanisms are proposed to generate the similarity weights. The ambiguous examples and their similarity-weights are thereafter incorporated into an SVM-based learning phase to build a more accurate classifier. Extensive experiments on real-world datasets have shown that SPUL outperforms state-of-the-art PU learning methods.

Semi-supervised learning, which uses a small amount of labeled data in conjunction with a large amount of unlabeled data for training, has recently attracted huge research attention due to the considerable improvement in learning accuracy. In this work, we focus on semi-supervised variable weighting for clustering, which is a critical step in clustering as it is known that interesting clustering structure usually occurs in a subspace defined by a subset of variables. Besides exploiting both labeled and unlabeled data to effectively identify the real importance of variables, our method embeds variable weighting in the process of semi-supervised clustering, rather than calculating variable weights separately, to ensure the computation efficiency. Our experiments carried out on both synthetic and real data demonstrate that semi-supervised variable weighting significantly improves the clustering accuracy of existing semi-supervised k-means without variable weighting, or with unsupervised variable weighting.

Enterprise data mining applications often involve complex data such as multiple large heterogeneous data sources, user preferences, and business impact. In such situations, a single method or one-step mining is often limited in discovering informative knowledge. It would also be very time and space consuming, if not impossible, to join relevant large data sources for mining patterns consisting of multiple aspects of information. It is crucial to develop effective approaches for mining patterns combining necessary information from multiple relevant business lines, catering for real business settings and decision-making actions rather than just providing a single line of patterns. The recent years have seen increasing efforts on mining more informative patterns, e.g., integrating frequent pattern mining with classifications to generate frequent pattern-based classifiers. Rather than presenting a specific algorithm, this paper builds on our existing works and proposes combined mining as a general approach to mining for informative patterns combining components from either multiple data sets or multiple features or by multiple methods on demand. We summarize general frameworks, paradigms, and basic processes for multifeature combined mining, multisource combined mining, and multimethod combined mining. Novel types of combined patterns, such as incremental cluster patterns, can result from such frameworks, which cannot be directly produced by the existing methods. A set of real-world case studies has been conducted to test the frameworks, with some of them briefed in this paper. They identify combined patterns for informing government debt prevention and improving government service objectives, which show the flexibility and instantiation capability of combined mining in discovering informative knowledge in complex data.

Adaptive Local Hyperplane (ALH) is a recently proposed classifier for the multi-class classification problems and it has shown encouraging performance in many pattern recognition problems. However, ALH's performance over many general classification datasets has only been tested by using a single loop of cross-validation procedure, where the whole datasets are used for both hyper-parameter determination and accuracy estimation. This procedure is appropriate for classifier performance comparison, but the produced results are likely to be optimistic for classifier accuracy estimation on new datasets. In this paper, we test the performance of ALH as well as several other benchmark classifiers by using two loops of cross-validation (a.k.a. double resampling) procedure, where the inner loop is used for hyper-parameter determination and the outer loop is used for accuracy estimation. With such a testing scheme, the classification accuracy of a tested classifier can be evaluated in a more strict way. The experimental results indicate the superior performance of the ALH classifier with respect to the traditional classifiers including Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Linear Discriminant Analysis (LDA), Classification Tree (Tree) and K-local Hyperplane distance Nearest Neighbor (HKNN). These results imply that the ALH classifier might become a useful tool for the pattern recognition tasks.

Most data mining algorithms and tools stop at the mining and delivery of patterns satisfying expected technical interestingness. There are often many patterns mined but business people either are not interested in them or do not know what follow-up actions to take to support their business decisions. This issue has seriously affected the widespread employment of advanced data mining techniques in greatly promoting enterprise operational quality and productivity. In this paper, we present a formal view of actionable knowledge discovery (AKD) from the system and decision-making perspectives. AKD is a closed optimization problem-solving process from problem definition, framework/model design to actionable pattern discovery, and is designed to deliver operable business rules that can be seamlessly associated or integrated with business processes and systems. To support such processes, we correspondingly propose, formalize, and illustrate four types of generic AKD frameworks: Postanalysis-based AKD, Unified-Interestingness-based AKD, Combined-Mining-based AKD, and Multisource Combined-Mining-based AKD (MSCM-AKD). A real-life case study of MSCM-based AKD is demonstrated to extract debt prevention patterns from social security data. Substantial experiments show that the proposed frameworks are sufficiently general, flexible, and practical to tackle many complex problems and applications by extracting actionable deliverables for instant decision making.

The seven papers in this special section focus on domain-driven data mining. One paper addresses the general framework of domain-driven data mining, four papers cover the approaches, and two papers consider the workable applications.

In this paper, we formulate a new research problem of learning from vaguely labeled one-class data streams, where the main objective is to allow users to label instance groups, instead of single instances, as positive samples for learning. The batch-labeling, however, raises serious issues because labeled groups may contain non-positive samples, and users may change their labeling interests at any time. To solve this problem, we propose a Vague One-Class Learning (VOCL) framework which employs a double weighting approach, at both instance and classifier levels, to build an ensembling framework for learning. At instance level, both local and global filterings are considered for instance weight adjustment. Two solutions are proposed to take instance weight values into the classifier training process. At classifier level, a weight value is assigned to each classifier of the ensemble to ensure that learning can quickly adapt to users' interests. Experimental results on synthetic and real-world data streams demonstrate that the proposed VOCL framework significantly outperforms other methods for vaguely labeled one-class data streams.

Named entity recognition aims at extracting named entities from unstructured text. A recent trend of named entity recognition is finding approximate matches in the text with respect to a large dictionary of known entities, as the domain knowledge encoded in the dictionary helps to improve the extraction performance.

Trading agents are useful for developing and back-testing quality trading strategies for taking actions in the real world. The existing trading agent research mainly focuses on simulation using artificial data. As a result, the actionable capability of developed trading strategies is often limited, and the trading agents therefore lack power. Actionable trading strategies can empower trading agents with workable decisionmaking in real-life markets. The development of actionable strategies is a non-trivial task, which needs to consider real-life constraints and organisational factors in the market. In this talk, we first analyse such constraints on developing actionable trading strategies for trading agents and propose a trading strategy development framework for trading agents. We then develop a series of trading strategies for trading agents through optimising, enhancing and discovering actionable trading strategies. We demonstrate working case studies using agent mining technology in real market data. These approaches, and their performance, are evaluated from both technical and business perspectives. These evalualtions clearly show that the development of trading strategies for trading agents, using our approach, can lead to smart decisions for brokerage firms and financial companies.

In SVMs-based multiple classification, it is not always possible to find an appropriate kernel function to map all the classes from different distribution functions into a feature space where they are linearly separable from each other. This is even worse if the number of classes is very large. As a result, the classification accuracy is not as good as expected. In order to improve the performance of SVMs-based multi-classifiers, this paper proposes a method, named multi-space-mapped SVMs, to map the classes into different feature spaces and then classify them. The proposed method reduces the requirements for the kernel function. Substantial experiments have been conducted on one-against-all, one-against-one, FSVM, DDAG algorithms and our algorithm using six UCI data sets. The statistical results show that the proposed method has a higher probability of finding appropriate kernel functions than traditional methods and outperforms others.

In this paper, we identify a new research problem on cleansing noisy data streams which contain incorrectly labeled training examples. The objective is to accurately identify and remove mislabeled data, such that the prediction models built from the cleansed streams can be more accurate than the ones trained from the raw noisy streams. For this purpose, we first use bias-variance decomposition to derive a maximum variance margin (MVM) principle for stream data cleansing. Following this principle, we further propose a local and global filtering (LgF) framework to combine the strength of local noise filtering (within one single data chunk) and global noise filtering (across a number of adjacent data chunks) to identify erroneous data. Experimental results on six data streams (including two real-world data streams) demonstrate that LgF significantly outperforms simple methods in identifying noisy examples.

Traditional sequential pattern mining deals with positive sequential patterns only, that is, only frequent sequential patterns with the appearance of items are discovered. However, it is often interesting in many applications to find frequent sequential patterns with the nonoccurrence of some items, which are referred to as negative sequential patterns. This paper analyzes three types of negative sequential rules and presents a new technique to find event-oriented negative sequential rules. Its effectiveness and efficiency are shown in our experiments.

Impact-targeted activities are rare but lead to significant impact on the society, e.g., isolated terrorism activities may lead to a disastrous event threatening national security. Similar issues can also be seen in many other areas. Therefore, it is important to identify such particular activities before they lead to significant impact to the world. However, it is challenging to mine impact-targeted activity patterns due to its imbalanced structure. This paper develops techniques for discovering such activity patterns. First, the complexities of mining imbalanced impact-targeted activities are analyzed.We then discuss strategies for constructing impact-targeted activity sequences. Algorithms are developed to mine frequent positive-impact (P rarr T) and negative-impact (P rarr Tmacr macr) oriented activity patterns, sequential impact-contrasted activity patterns (P is frequently associated with both pattern P rarr T and P rarr Tmacr macr in separated data sets), and sequential impact-reversed activity patterns (both P rarr T and PQ rarr T macr are frequent). Activity impact modelling is also studied to quantify pattern impact on business outcomes. Social security debt-related activity data is used to test the proposed approaches. The outcomes show that they are promising for ISI applications to identify impact-targeted activity patterns in imbalanced data.

Open complex agent systems (OCAS) are becoming increasingly important in constructing problem-solving systems for enterprise applications. are challenging because they present very high system complexities involving human users and interactions with a changing environment. The existing agent-oriented software engineering (AOSE) approaches have trouble in engineering OCAS because of a number of deficiencies, e.g., lacking the capability of handling system dynamics analysis. This paper introduces an effective AOSE approach, i.e., organization- and service-oriented system analysis and design (OSOAD). It is used to extract and model system members and design a real-life OCAS system called financial trading rule automated development and evaluation (F-Trade). Through the case studies of visual and formal modeling and design of major organizational members, relations, and subsystems in F-Trade, this paper demonstrates the effective mechanisms and capabilities of the OSOAD approach. System implementation and evaluation results further show that OSOAD provides comprehensive AOSE support for engineering real-world open complex agent organizations.

Traditionally, data mining is an autonomous data-driven trial-and-error process. Its typical task is to let data tell a story disclosing hidden information, in which domain intelligence may not be necessary in targeting the demonstration of an algorithm. Often knowledge discovered is not generally interesting to business needs. Comparably, real-world applications rely on knowledge for taking effective actions. In retrospect of the evolution of KDD, this paper briefly introduces domain-driven data mining to complement traditional KDD. Domain intelligence is highlighted towards actionable knowledge discovery, which involves aspects such as domain knowledge, people, environment and evaluation. We illustrate it through mining activity patterns in social security data.

A bridging rule in this paper has its antecedent and action from different conceptual clusters. We first design two algorithms for mining bridging rules between clusters in a database, and then propose two non-linear metrics for measuring the interestingness of bridging rules. Bridging rules can be distinct from association rules (or frequent itemsets). This is because (1) bridging rules can be generated by infrequent itemsets that are pruned in association rule mining; and (2) bridging rules are measured by the importance that includes the distance between two conceptual clusters, whereas frequent itemsets are measured by only the support.

Early requirements analysis (ERA) is quite significant for building agent-based systems. Goal-oriented requirements analysis is promising for the agent-oriented early requirements analysis. In general, either visual modeling or formal specifications is used for the ERA. This way cannot capture requirements precisely and completely. We present an integrative modeling framework for agent-oriented early requirements analysis; this framework implements goal-oriented requirement analysis. The integrative modeling combines visual modeling and formal modeling together. Extended i* framework is used for building visual models; formal specifications complement the visual modeling to define and refine requirements. Both visual and formal models are outlined through a practical agent-based system F-TRADE. The integrative modelling seems to model early requirements comprehensively and concretely, and benefit refinement and conflict management in building agent systems.

Usually, integration of business intelligence (BI) from realistic telecom enterprise is by packing data warehouse (DW), OLAP, data mining and reporting from different vendors together. As a result, BI system users are transferred to a reporting system with reports, data models, dimensions and measures predefined by system designers. As a result of survey, 85% of DW projects failed to meet their intended objectives. In this paper, we investigate how to integrate BI packages into an adaptive and flexible knowledge portal by constructing an internal link and communication channel from top-level business concepts to underlying enterprise information systems (EIS). An approach of three-level ontology services is developed, which implements unified naming, directory and transport of ontology services, and ontology mapping and query parsing among conceptual view, analytical view and physical view from user interfaces through DW to EIS. Experiments on top of real telecom EIS shows that our solution for integrating BI presents much stronger power to support operational decision making more user-friendly and adaptively compared with those simply combining BI products presently available together.

Density-based and grid-based clustering are two main clustering approaches. The former is famous for its capability of discovering clusters of various shapes and eliminating noises, while the latter is well known for its high speed. Combination of the two approaches seems to provide better clustering results. To the best of our knowledge, however, all existing algorithms that combine density-based clustering and grid-based clustering take cells as atomic units, in the sense that either all objects in a cell belong to a cluster or no object in the cell belong to any cluster. This requires the cells to be small enough to ensure the fine resolution of results. In high-dimensional spaces, however, the number of cells can be very large when cells are small, which would make the clustering process extremely costly. On the other hand, the number of neighbors of a cell grows exponentially with the dimensionality of datasets, which makes the complexity increase further. In this paper, we present a new approach that takes objects (or points) as the atomic units, so that the restriction of cell size can be relaxed without degrading the resolution of clustering results. In addition, a concept of ith-order neighbors is introduced to avoid considering the exponential number of neighboring cells. By considering only low-order neighbors, our algorithm is very efficient while losing only a little bit of accuracy. Experiments on synthetic and public data show that our algorithm can cluster high-dimensional data effectively and efficiently.

Today's database systems must deal with uncertainty in the data they store. Consequently, there is a strong need for mining probabilistic databases. Because probabilistic data in first normal form relations is redundant, existing mining techniques are inadequate for discovering probabilistic databases. This paper designs a new strategy for identifying potentially useful patterns in probabilistic databases. A dependent rule is thus identified in a probabilistic database, represented in the form X → Y with conditional probability matrix MY|X . This method uses an instance selection to increase efficiency, enabling us to reduce the search space. We evaluated the proposed technique, and our experimental results demonstrate that the approach is effective and efficient.

Identifying software component association is useful for component management and component retrieval. In this paper we design an evolutionary strategy to understand software structure better and identify software component association, by using genetic algorithm. Our mining strategy is effective for global search, especially when the searched space is so large that it is hardly possible to use deterministic search method.

This paper presents an efficient method for mining both positive and negative association rules in databases. The method extends traditional associations to include association rules of forms A ⇒ ¬ B , ¬ A ⇒ B , and ¬ A ⇒ ¬ B , which indicate negative associations between itemsets. With a pruning strategy and an interestingness measure, our method scales to large databases. The method has been evaluated using both synthetic and real-world databases, and our experimental results demonstrate its effectiveness and efficiency.

For making good decisions in the area of petroleum production, it is becoming a big problem how to timely gather sufficient and correct information, which may be stored in databases, data files, or on the World Wide Web. In this paper, Gaia methodology and Open Agent Architecture were employed to contribute a framework to solve above problem. The framework consists of three levels, namely, role mode, agent type, and agent instance. The model with five roles is analyzed. Four agent types are designed Six agent instances are developed for constructing the system of petroleum information services. The experimental results show that all agents in the system can work cooperatively to organize and retrieve relevant petroleum information. The successful implementation of the framework shows that agent-based technology can significantly facilitate the construction of complex systems in distributed heterogeneous data resource environment.

Presents an efficient and effective application-independent database cluster for mining multi-databases. Our experimental results show that the proposed approach is effective and promising.

This paper develops a hybrid model which provides a unified framework for the following four kinds of reasoning: 1) Zadeh's fuzzy approximate reasoning; 2) truth-qualification uncertain reasoning with respect to fuzzy propositions; 3) fuzzy default reasoning (proposed, in this paper, as an extension of Reiter's default reasoning); and 4) truth-qualification uncertain default reasoning associated with fuzzy statements (developed in this paper to enrich fuzzy default reasoning with uncertain information). Our hybrid model has the following characteristics: 1) basic uncertainty is estimated in terms of words or phrases in natural language and basic propositions are fuzzy; 2) uncertainty, linguistically expressed, can be handled in default reasoning; and 3) the four kinds of reasoning models mentioned above and their combination models will be the special cases of our hybrid model. Moreover, our model allows the reasoning to be performed in the case in which the information is fuzzy, uncertain and partial. More importantly, the problems of sharing the information among heterogeneous fuzzy, uncertain and default reasoning models can be solved efficiently by using our model. Given this, our framework can be used as a basis for information sharing and exchange in knowledge-based multi-agent systems for practical applications such as automated group negotiations. Actually, to build such a foundation is the motivation of this paper.

Database systems have been designed to serve multi-users in real-world applications. There are essential differences between mono- and multi-user applications when a database is very large. Therefore, this paper presents an "anytime" framework for mining very large databases which are shared by multi-users. Anytime mining is designed to generate approximate results such that these results can be accessed at any time while the system is autonomously mining a database.

Complexity reduction is an important task in Bayesian networks. Recently, an approach known as the linear potential function (LPF) model has been proposed for approximating Bayesian computations. The LPF model can effectively compress a conditional probability table into a linear function. This correspondence extends the LPF model to approximate propagation in Bayesian networks. The extension focuses on encoding probability propagation as a polynomial function for a class of tractable problems.

Current data mining techniques may not be helpful for mining some companies/organizations such as nuclear power plants and earthquake bureaus, which have only small databases. Apparently, these companies/organizations also expect to apply data mining techniques to extract useful patterns in their databases so as to make their decisions. However, data in these databases such as the accident database of a nuclear power plant and the earthquake database in an earthquake bureau, may not be large enough to form any patterns. To meet the applications, we present a new mining model in this paper, which is based on the collecting knowledge from such as Web, journals, and newspapers.

Many applications such as marketing and stock investment may be time limited and few considerations to accurate itemsets. For these applications, we present a mining model to quickly estimate the approximate support of frequent itemsets of interest in large scale databases in this paper. An efficient algorithm is thus designed to reduce the searched space by pruning uninterested frequent itemsets.

In this paper, we prove that, under the assumption of conditional independence, the EMYCIN formula for sequential propagation can be derived strictly from the definition of the certainty factor according to probability theory. J.B. Adams (1984) and S. Schocken (1988) have proved that the EMYCIN formula for parallel propagation is partially consistent with probability theory. Our result supplements their contributions and, together with theirs, explains why the EMYCIN certainty factor model works reasonably well.

In this paper, firstly, potential synthesis cases in distributed expert systems (DESs) and types of DESs are identified. Based on these results, necessary conditions of synthesis strategies in different synthesis cases are recognized. Secondly, two methodologies for designing synthesis strategies in DESs are investigated. They are analysis methods and inductive methods. Thirdly, two methodologies are discussed based on the viewpoints of performance, complexity and requirements.

The task of diagnosis, a typical abductive problem, as to find a hypothesis that best explains a set of observations. Generally, a neural network diagnostic reasoning model finds only one hypothesis to a set of observations. It is computationally expensive to find the hypothesis because the number of the potential hypotheses is exponentially large. Recently, we have proposed a connectionist diagnosis model to overcome the above difficulty. In this paper, we propose a method to improve the efficiency and the practicality of the model. The improved model can find more solutions, and the efficiency of the model is also improved.

The "take-them-in-or-leave-them-out" of prior probabilities is a key problem in uncertain reasonings. The EMYCIN uncertain reasoning model is inconsistent with probability theory, due to 'leaving them out', while the PROSPECTOR uncertain reasoning model is substantially consistent with probability theory, due to 'taking them in'. This presents a difficult task for human experts when supplying prior probabilities. In this paper, in order to overcome the difficulty, we propose a hybrid uncertain reasoning model, by combing rule strengths in the PROSPECTOR model with rule strengths in the EMYCIN model. Moreover, different forms of rule strength will enable human experts to supply values for the rule strengths more flexibly in a knowledge base. Finally, an example is given to illustrate the power of our methodology.

Model-based diagnosis has the disadvantage of a high computational complexity. One way to overcome this disadvantage is to focus the diagnosis on a reduced diagnostic space. We propose an improved critical diagnosis reasoning method based on the method proposed by (Raiman et al., 1993). The method focuses the diagnosis on finding out the kernel diagnoses instead of the whole diagnoses. We give an updated definition of critical cover which we call "critical partition". The conditions satisfied by critical partition are relaxed compared with the conditions for critical cover. Correspondingly, a non-backtracking algorithm called Searching Critical Partition (SCP) to find out the critical partition is also proposed.

This paper presents the design and implementation of a knowledge-based communication system for distributed expert systems. This communication system has been designed to handle the wide range of protocols among cooperating expert systems, to be flexible to cover new application domains, and to be efficient in its implementation. It has been implemented using a specialized forward-chaining rule-based system in order to achieve these features and applied in a prototype framework for distributed expert systems, called HECODES.>

Cooperation between different expert systems within a heterogeneous distributed expert system is essentially based on the transformations between the reasoning models built into the structures of the involved systems. The general criteria for a transformation between heterogeneous inexact reasoning models have been formulated. Using an algebraic structure of semigroup (with individual unit elements) of inexact reasoning models, the criteria are natural and easily satisfied by homomorphic mapping as an implementation. The criteria presented are too weak to ensure the existence of a unique mapping between the models concerned, but they are strong enough to guarantee reasonable transformations. These general considerations are illustrated using the example of three inexact reasoning models used by EMYCIN, PROSPECTOR, and MYCIN, where the homomorphic transformations among them are constructed and discussed.>

The causes of deadlocks in distributed expert systems are examined and classified, and two methods of processing them are proposed. In particular, this problem is discussed within the framework of a working prototype distributed expert system, Hecodes (heterogeneous cooperative distributed expert system) which consists of a control system, a blackboard system, and a group of expert systems whose structures and functions can be different. The two approaches for processing any deadlock are termed deadlock avoidance, and deadlock detection and destruction. Only one of the two can be chosen for a particular run of the system. The detailed working and effects of using each of the proposed deadlock resolution methods are discussed.>