Huan Li

Time series anomaly detection (TSAD) has long been a hot research topic in data mining due to its various applications. Recent studies challenge the effectiveness of popular deep learning methods for TSAD, suggesting their failure in detecting subtle and prolonged anomalies. Outlier Exposure (OE) and Masked Autoencoder (MAE) emerge as two promising paradigms (classification and reconstruction) for solving the above problems. However, OE-based methods are constrained by poor generalization, while MAE-based methods are limited by masking misalignment issues. To address these limitations, this paper proposes a novel framework, CoAD, which unifies the two paradigms to leverage their complementary strengths while mitigating their respective weaknesses. In this framework, the classification module generates probability-informed soft masks for the reconstruction module, which in turn alleviates the generalization problem of the classification module. This cooperative design enables CoAD to effectively detect subtle and complex anomalies that are often overlooked by existing methods. Additionally, the classification module is carefully designed to resolve issues related to improper classification granularity and the neglect of frequency information. Extensive experiments on high-quality benchmark datasets, conducted under rigorous evaluation protocols, demonstrate that CoAD significantly outperforms both state-of-the-art deep learning and traditional data mining methods, highlighting the potential of deep learning in TSAD. Moreover, CoAD is lightweight and substantially faster than existing SOTA methods, demonstrating its practical value for large-scale, real-time applications.

Machine learning (ML) on tabular data is ubiquitous, yet obtaining abundant high-quality tabular data for model training remains a significant obstacle. Numerous works have focused on tabular data augmentation (TDA) to enhance the original table with additional data, thereby improving downstream ML tasks. Recently, there has been a growing interest in leveraging the capabilities of generative AI for TDA. Therefore, we believe it is time to provide a comprehensive review of the progress and future prospects of TDA, with a particular emphasis on the trending generative AI. Specifically, we present an architectural view of the TDA pipeline, comprising three main procedures: pre-augmentation, augmentation, and post-augmentation. Pre-augmentation summarizes eight commonly-used preparation tasks that facilitate subsequent TDA from different perspectives. Augmentation systematically analyzes current TDA methods, categorized into retrieval-based methods, which retrieve external data, and generation-based methods, which generate synthetic data. We further subdivide these methods based on the augmentation process at the row, column, cell, and table levels. Post-augmentation focuses on the datasets, evaluation and optimization aspects of TDA. We also summarize current trends and future directions for TDA, highlighting promising opportunities in the era of generative AI. The accompanying articles and related resources are continuously updated in GitHub.

Session-based recommendation (SBR), which focuses on next-item prediction for anonymous users based on short-term interaction sequences, has garnered increasing attention from researchers. While graph neural networks (GNNs) have become predominant in modeling complex item transition patterns, our empirical study reveals two critical limitations in existing GNN-based SBR methods. On the one hand, they struggle to differentiate between sessions sharing the same last item, resulting in indistinguishable session representations. On the other hand, the inherent popularity bias in session data leads to the over-recommendation of popular items. Inspired by contrastive learning techniques, this paper presents a unified mitigation framework for Same lAst-item confusion in Graph lEarning (SAGE) for SBR. In SAGE, we first obtain normalized session embeddings on constructed session graphs. We then build positive and negative samples of sessions through dual forward propagations and a novel negative sample selection strategy, followed by calculating contrastive loss. Finally, the enhanced session embeddings are utilized for prediction. Extensive experiments on two real-world datasets demonstrate that integrating SAGE with various state-of-the-art GNN-based SBR methods significantly improves their original performances.

Next Point-of-interest (POI) recommendation has been widely used in real scenarios to predict the next possible location based on user behavior patterns. However, existing methods predominantly rely on spatio-temporal associations and check-in sequence relationships between users and POIs, which fall short for users with limited interactions with POIs. Moreover, user preferences are inherently multi-dimensional, rendering user selections often influenced by multiple factors such as location categories and multi-modal information. To mitigate these issues, we introduce a Multi-Modal Knowledge Graph Modeling of Multi-Dimensional User Preferences for Next-POI Recommendation (M4Rec for short). First, we define a multi-modal knowledge graph to organize the relationships among users, locations, categories, and multi-modal information. Subsequently, we use the multi-modal knowledge graph-based relation-aware network to derive comprehensive entity representations from the constructed knowledge graph. Next, employing the temporal knowledge prediction method, we predict the user's next-POI category and next-POI. Finally, the final recommendation results are obtained by enhancing the corresponding location prediction scores through category semantics. Extensive experimentation conducted on real-world datasets validates the superiority of our proposed method over state-of-the-art competitors.

While Vision-Language Models (VLMs) have shown promise in textual understanding, they face significant challenges when handling long context and complex reasoning tasks. In this paper, we dissect the internal mechanisms governing long-context processing in VLMs to understand their performance bottlenecks. Through the lens of attention analysis, we identify specific Visual Evidence Retrieval (VER) Heads - a sparse, dynamic set of attention heads critical for locating visual cues during reasoning, distinct from static OCR heads. We demonstrate that these heads are causal to model performance; masking them leads to significant degradation. Leveraging this discovery, we propose VERA (Visual Evidence Retrieval Augmentation), a training-free framework that detects model uncertainty (i.e., entropy) to trigger the explicit verbalization of visual evidence attended by VER heads. Comprehensive experiments demonstrate that VERA significantly improves long-context understanding of open-source VLMs: it yields an average relative improvement of 21.3% on Qwen3-VL-8B-Instruct and 20.1% on GLM-4.1V-Thinking across five benchmarks.

Fashion image editing has garnered significant attention due to its growing demand in e-commerce, social media, and virtual try-on applications. However, existing methods are typically designed for specific editing tasks in isolation, lacking a unified framework capable of handling diverse editing requirements. This work addresses this limitation from two critical perspectives. First, we construct InstructFashion, a large-scale, high-quality dataset specifically curated for instruction-guided fashion image editing. It is generated through carefully designed pipelines that cover four distinct editing tasks. Second, we propose TailorEdit, an adaptive framework for instruction-guided fashion image editing. It integrates human segmentation map-based denoising guidance, modular LoRA-based editing experts, and a dynamic expert routing mechanism to enable precise and semantically coherent modifications. Extensive quantitative and qualitative evaluations demonstrate that TailorEdit consistently outperforms state-of-the-art methods in terms of realism, coherence, and instruction adherence. Our code is available at https://github.com/EndaJude/TailorEdit.

With streaming floating-point numbers being increasingly prevalent, effective and efficient compression of such data is critical. Compression schemes must be able to exploit the similarity, or smoothness, of consecutive numbers and must be able to contend with extreme conditions, such as high-precision values or the absence of smoothness. We present DeXOR, a novel framework that enables decimal XOR procedure to encode decimal-space longest common prefixes and suffixes, achieving optimal prefix reuse and effective redundancy elimination. To ensure accurate and low-cost decompression even with binary-decimal conversion errors, DeXOR incorporates 1) scaled truncation with error-tolerant rounding and 2) different bit management strategies optimized for decimal XOR. Additionally, a robust exception handler enhances stability by managing floating-point exponents, maintaining high compression ratios under extreme conditions. In evaluations across 22 datasets, DeXOR surpasses state-of-the-art schemes, achieving a 15% higher compression ratio and a 20% faster decompression speed while maintaining a competitive compression speed. DeXOR also offers scalability under varying conditions and exhibits robustness in extreme scenarios where other schemes fail.

Deep neural network (DNN) models are crucial for Internet-of-Things (IoT) applications. In a multi-access edge computing (MEC) system, IoT devices dispatch their requests to a nearby edge server to accelerate DNN inference. Recently, processing-in-memory (PIM) accelerators have emerged for DNN inference due to their power and area efficiency. However, current scheduling methods primarily focus on single-DNN workloads and do not fully utilize PIM’s potential for managing multi-DNN workloads in MEC scenarios. Additionally, PIM accelerators face challenges such as inflexible resource allocation and costly write operations, constraining the direct application of GPU scheduling techniques. In this paper, we define the multi-DNN scheduling problem on the PIM accelerator. To reduce the problem’s complexity, we propose a heuristic optimization algorithm, called Leader-Follower. Using this algorithm, we present PimShare, a scheduler for the PIM accelerator to process multi-DNN workloads. PimShare schedules multi-DNN inference requests to enable both temporal and spatial multiplexing, which can enhance hardware utilization. Compared with baseline scheduling methods, including those adapted from GPU scheduling, PimShare can achieve up to a two-order-of-magnitude improvement in throughput and scale the high throughput to workloads containing 18 DNN models. In addition, PimShare reduces write operations and processing latency.

With the proliferation of GPS-equipped edge devices, huge trajectory data are generated and accumulated in various domains, driving numerous urban applications. However, due to the limited data acquisition capabilities of edge devices, many trajectories are often recorded at low sampling rates, reducing the effectiveness of these applications. To address this issue, we aim to recover high-sample-rate trajectories from low-sample-rate ones enhancing the usability of trajectory data. Recent approaches to trajectory recovery often assume centralized data storage, which can lead to catastrophic forgetting, where previously learned knowledge is entirely forgotten when new data arrives. This not only poses privacy risks but also degrades performance in decentralized settings where data streams into the system incrementally. To enable decentralized training and streaming trajectory recovery, we propose a Lightweight incremental framework for federated Trajectory Recovery, called LightTR+, which is based on a client-server architecture. Given the limited processing capabilities of edge devices, LightTR+ includes a lightweight local trajectory embedding module that enhances computational efficiency without compromising feature extraction capabilities. To mitigate catastrophic forgetting, we propose an intra-domain knowledge distillation module. Additionally, LightTR+ features a meta-knowledge enhanced local-global training scheme, which reduces communication costs between the server and clients, further improving efficiency. Extensive experiments offer insight into the effectiveness and efficiency of LightTR+.