Information Sciences最新文献_第8页

Data stream clustering via fuzzy similarity and diffusion-enhanced contextual affinity 基于模糊相似性和扩散增强的上下文亲和力的数据流聚类

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122690

Yao Li , Ming Chi , Wei Lu , Xiaodong Liu , Witold Pedrycz

{"title":"Data stream clustering via fuzzy similarity and diffusion-enhanced contextual affinity","authors":"Yao Li , Ming Chi , Wei Lu , Xiaodong Liu , Witold Pedrycz","doi":"10.1016/j.ins.2025.122690","DOIUrl":"10.1016/j.ins.2025.122690","url":null,"abstract":"<div><div>Data stream clustering provides an effective method for recognizing underlying patterns in potentially unbounded sequences of data objects. Existing data stream clustering methods primarily encounter two key issues: (1) the inadequate evaluation of relationships between data objects within fixed-size landmark windows, leading to degraded clustering quality; and (2) the absence of efficient mechanisms for transferring useful knowledge from previous windows to the current window, weakening the model’s adaptability to data stream evolution. To address these issues, a data stream clustering method based on axiomatic fuzzy set theory via a diffusion process is first proposed. First, the proposed method employs axiomatic fuzzy set theory to measure the relationships between data objects within the window, capturing similarity information to more accurately reveal the underlying data distribution. Second, an efficient diffusion process enhances pairwise affinities through contextual propagation, which significantly improves connectivity within clusters. Finally, the learned affinity matrix is applied to spectral clustering for data stream clustering. Over time, we update the dynamic set according to the distance between data objects and cluster centers. This dynamic set retains representative data objects and effectively transfers previously learned knowledge to the current landmark window. Experimental results on four datasets and seven algorithms demonstrate the effectiveness and robustness of the proposed method.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122690"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107116","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Existence of nash equilibrium in single-leader multiple-follower games with min-max interval payoffs 具有最小-最大区间收益的单领导多随从对策纳什均衡的存在性

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122691

Daping Zhang , Yanlong Yang , Xicai Deng

{"title":"Existence of nash equilibrium in single-leader multiple-follower games with min-max interval payoffs","authors":"Daping Zhang , Yanlong Yang , Xicai Deng","doi":"10.1016/j.ins.2025.122691","DOIUrl":"10.1016/j.ins.2025.122691","url":null,"abstract":"<div><div>In decision-making problems with hierarchical structures, leader–follower games are highly prevalent. As a core concept in game theory, the existence of Nash equilibrium is crucial. However, in reality, complex uncertainties often lead to imprecise game outcomes, and interval representations are an effective tool for capturing such uncertainties. To address the issue of imprecise payoffs in complex environments, this paper proposes the concept of min-max interval (for short, MMI) and studies the existence of Nash equilibrium in single-leader multiple-follower (for short, SLMF) games with MMI payoffs. MMI is an appropriate extension of the traditional interval-providing a more flexible tool for representing uncertain payoffs. We propose an MMI expected payoff ranking method to address the issue of players ranking MMIs. Based on this, operational rules for MMIs and concepts such as limits, continuity, and concavity of MMI-valued functions (for short, MIVFs) are defined. After extending key theorems of real-valued functions to the case of MIVFs, we combine these extended theorems with set-valued mapping theory and Kakutani’s fixed point theorem to prove the existence of Nash equilibrium in SLMF MMI-valued games. Additionally, we compare existing works to verify the innovativeness of the proposed method and provide numerical examples to demonstrate its applicability.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122691"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep dual contrastive learning for multi-view subspace clustering 多视图子空间聚类的深度对偶对比学习

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122678

Xincan Lin , Jie Lian , Zhihao Wu , Jielong Lu , Shiping Wang

{"title":"Deep dual contrastive learning for multi-view subspace clustering","authors":"Xincan Lin , Jie Lian , Zhihao Wu , Jielong Lu , Shiping Wang","doi":"10.1016/j.ins.2025.122678","DOIUrl":"10.1016/j.ins.2025.122678","url":null,"abstract":"<div><div>Multi-view subspace clustering (MVSC) aims to learn a consistent shared self-representation by utilizing the consistency and complementarity of all views, numerous MVSC algorithms have attempted to obtain the optimal representation directly from raw features. However, they might overlook the noisy or redundant information in raw feature space, resulting in learning suboptimal self-representation and poor performance. To address this limitation, an intuitive idea is introducing deep neural networks to eliminate the noise and redundancy, yielding a potential embedding space. Nevertheless, existing deep MVSC methods merely focus on either the embeddings or self-expressions to explore the complementary information, which hinders subspace learning. In this paper, we present a deep multi-view dual contrastive subspace clustering framework to exploit the complementarity to learn latent self-representations effectively. Specifically, multi-view encoders are constructed to eliminate noise and redundancy of the original features and capture low-dimensional subspace embeddings, from which the self-representations are learned. Moreover, two diverse specific fusion methods are conducted on the latent subspace embeddings and the self-expressions to learn shared self-representations, and dual contrastive constraints are proposed to fully exploit the complementarity among views. Extensive experiments are conducted to verify the effectiveness of the proposed method.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122678"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107570","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Robust watermarking for diffusion model generated images 扩散模型生成图像的鲁棒水印

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122686

Ziqi Liu , Yuan Guo , Liansuo Wei

{"title":"Robust watermarking for diffusion model generated images","authors":"Ziqi Liu , Yuan Guo , Liansuo Wei","doi":"10.1016/j.ins.2025.122686","DOIUrl":"10.1016/j.ins.2025.122686","url":null,"abstract":"<div><div>With the wide application of diffusion models in the field of image generation, image copyright protection and traceability have become increasingly complex and challenging. To address these problems, this paper proposes a robust watermarking method for diffusion model generated images to achieve their copyright protection and traceability. The method designs an invertible mapping module to replicate and cryptographically map the watermark information into an approximately Gaussian distributed noise, which is highly consistent with the distribution of the original generation model. The mapped watermark noise serves as the latent space vector of the generative model, preserving both image generation quality and model performance. In the watermark extraction stage, the original watermark information can be accurately recovered from the generated image through the reverse extraction and voting mechanism. Experimental results show that the proposed method demonstrates excellent performance in terms of image watermark extraction accuracy, robustness and watermark image generation quality. It can still maintain 99 % true positive rate and 97.5 % bit accuracy under various attacks, and the overall performance in the detection and traceability scenarios is significantly better than the existing baseline methods.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122686"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-branch semantic alignment for few-shot image classification 基于多分支语义对齐的少镜头图像分类

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122676

Zijun Zheng , Heng Wu , Laishui Lv , Changchun Zhang , Hongcheng Guo , Shanzhou Niu , Gaohang Yu

{"title":"Multi-branch semantic alignment for few-shot image classification","authors":"Zijun Zheng , Heng Wu , Laishui Lv , Changchun Zhang , Hongcheng Guo , Shanzhou Niu , Gaohang Yu","doi":"10.1016/j.ins.2025.122676","DOIUrl":"10.1016/j.ins.2025.122676","url":null,"abstract":"<div><div>The remarkable progress of deep learning in computer vision has significantly stimulated research interest in few-shot image classification. This field aims to transfer knowledge from previous experiences to recognize new concepts with limited samples. However, most existing approaches primarily concentrate on aligning semantic information at high-level features, neglecting the importance of middle-level or low-level feature representations. In this paper, we propose a novel approach called Multi-Branch Semantic Alignment (MBSA) for few-shot image classification, with the objective of investigating the role of multi-level features. Instead of using standard convolutional layers, we employ diverse convolutional layers to generate enhanced representations in each branch. These representations are then utilized by a dense classifier, which is supervised by a powerful guidance mechanism to incorporate semantic information into their spatial locations. During the inference stage, the multi-branch semantic alignment is designed to align multi-level features between query images and support images. This alignment process effectively establishes semantic correspondences between representations at different levels, thereby enhancing the ability to recognize novel categories. Comprehensive experiments are conducted on various few-shot benchmarks to demonstrate the superiority of our approach compared to those of several previous approaches, and ablation studies are performed to analyze the impact of different components.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122676"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107575","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SEAD-MGFE-Net: Schrödinger equation-based adaptive dropout multi-granular feature enhancement network for conversational aspect-based sentiment quadruple analysis Schrödinger基于方程的自适应dropout多颗粒特征增强网络，用于会话方面的情感四重分析

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-15 DOI: 10.1016/j.ins.2025.122684

Wei Liu , Xiaoliang Chen , Duoqian Miao , Hongyun Zhang , Xiaolin Qin , Shangyi Du , Peng Lu

{"title":"SEAD-MGFE-Net: Schrödinger equation-based adaptive dropout multi-granular feature enhancement network for conversational aspect-based sentiment quadruple analysis","authors":"Wei Liu , Xiaoliang Chen , Duoqian Miao , Hongyun Zhang , Xiaolin Qin , Shangyi Du , Peng Lu","doi":"10.1016/j.ins.2025.122684","DOIUrl":"10.1016/j.ins.2025.122684","url":null,"abstract":"<div><div>This research focuses on enhancing the extraction of sentiment quadruples consisting of target, aspect, opinion, and sentiment from multi-turn dialogs, which remains a challenging problem in conversational sentiment analysis. Existing methods frequently encounter challenges with complex sentence structures, presence of multiple sentiment quadruples, and interference from irrelevant contextual information. These challenges often result in suboptimal performance. These limitations are addressed by introducing Schrödinger equation-based adaptive dropout multi-granular feature enhancement network (SEAD-MGFE-Net), a novel framework that synergizes multigranular feature extraction with quantum-inspired adaptive regularization. The proposed methodology incorporates a multi-layer tree structure to segment sentences into semantically coherent fragments, thereby improving the alignment between aspect and opinion terms while simultaneously mitigating noise impact. Moreover, we engineer a multi-angle dynamic adjacency learning enhancement module that adeptly captures both local and global features inherent in graph-structured representations. Additionally, we devise an adaptive dropout mechanism based on the Schrödinger equation, facilitating automatic modulation of the regularization strength throughout training. Extensive evaluations on benchmark datasets in both Chinese and English validate the state-of-the-art effectiveness of our proposed SEAD-MGFE-Net model, achieving Micro-F1 scores of 46.53 % (Chinese) and 40.97 % (English), surpassing the strongest baseline models by 2.04 % and 1.57 %, respectively. SEAD-MGFE-Net exhibits efficacy in extracting cross-utterance quadruples and managing long-range dependencies. These findings confirm the effectiveness and broad applicability of SEAD-MGFE-Net for conversational sentiment analysis.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122684"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107120","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Spatio-frequency texture analysis using wavelet increment entropy: Methodology and application to MRI in multiple sclerosis 基于小波增量熵的空间频率纹理分析：方法及其在多发性硬化症MRI中的应用

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-12 DOI: 10.1016/j.ins.2025.122669

Muqaddas Abid , Muhammad Suzuri Hitam , Rozniza Ali , Hamed Azami , Anne Humeau-Heurtier

{"title":"Spatio-frequency texture analysis using wavelet increment entropy: Methodology and application to MRI in multiple sclerosis","authors":"Muqaddas Abid , Muhammad Suzuri Hitam , Rozniza Ali , Hamed Azami , Anne Humeau-Heurtier","doi":"10.1016/j.ins.2025.122669","DOIUrl":"10.1016/j.ins.2025.122669","url":null,"abstract":"<div><div>Texture analysis is crucial for understanding images by extracting features that define spatial patterns. Recently, bi-dimensional extensions of entropy measures have gained attention due to their simplicity and strong theoretical foundations. However, existing methods primarily operate in the spatial domain and thus overlook frequency-domain and multiscale information. To address this, we introduce bidimensional wavelet increment entropy (wavelet IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span>). A one-level discrete wavelet transform (DWT) with the Haar wavelet decomposes each image into approximation (low-frequency) and, for some neuroimaging data, detail (high-frequency) subbands; IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span> is then applied both to capture global structural patterns and fine, detailed texture variations. We evaluated wavelet IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span> on synthetic and real datasets, demonstrating its effectiveness in distinguishing between different noise types (white Gaussian, salt-and-pepper, and speckle noise). Comparisons between periodic and synthesized images revealed lower wavelet IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span> values for periodic textures. Tests on real texture datasets highlight the method's ability to differentiate various patterns. In particular, wavelet IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span> achieved 86.69% accuracy in distinguishing MRI images of healthy versus multiple sclerosis–affected brains. Overall, wavelet IncrEn<span><math><msub><mrow></mrow><mrow><mn>2</mn><mi>D</mi></mrow></msub></math></span> offers a robust, frequency-aware descriptor that outperforms existing 2D entropy methods.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122669"},"PeriodicalIF":6.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Parameter-free discrete clustering via adaptive hypergraph fusion 基于自适应超图融合的无参数离散聚类

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-12 DOI: 10.1016/j.ins.2025.122677

Yu Zhou , Ben Yang , Xuetao Zhang , Badong Chen

{"title":"Parameter-free discrete clustering via adaptive hypergraph fusion","authors":"Yu Zhou , Ben Yang , Xuetao Zhang , Badong Chen","doi":"10.1016/j.ins.2025.122677","DOIUrl":"10.1016/j.ins.2025.122677","url":null,"abstract":"<div><div>Graph-based clustering has garnered significant attention due to its outstanding performance in uncovering sample structures. However, existing graph-based methods face two major challenges: 1) In graph construction, they typically focus only on direct connections between samples or an exact high-order relationship, neglecting the impact of hidden complex relationships on clustering performance; 2) The separation of spectral analysis and category acquisition into two distinct stages often results in a loss of effectiveness. To handle these problems, we propose a parameter-free discrete clustering method, called parameter-free discrete clustering via adaptive hypergraph fusion (DCAHF). Specifically, DCAHF first produces multiple different hypergraphs, each serving as a biased approximation of the data's intrinsic manifold. These complementary approximations capture distinct local-to-global geometric patterns. Then, it introduces an adaptive fusion strategy that learns optimal weights to combine them into a single consensus hypergraph on manifold space, effectively reconstructing the real manifold structure with reduced bias and improved integrity. Finally, discrete spectral analysis is performed directly on the consensus hypergraph to generate discrete sample categories, thereby avoiding the performance loss associated with two-stage approaches. Thus, DCAHF is a high-performance, parameter-free clustering model that can flexibly adapt to various clustering tasks. Since the DCAHF model cannot be solved using gradient descent methods, we develop a coordinate descent-based optimization algorithm to efficiently solve the model. Extensive experimental results demonstrate that DCAHF significantly enhances clustering effectiveness while maintaining comparable efficiency to state-of-the-art methods.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122677"},"PeriodicalIF":6.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107572","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MPCMO: An improved multi-population co-evolutionary algorithm for many-objective optimization MPCMO：一种改进的多种群协同进化多目标优化算法

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-12 DOI: 10.1016/j.ins.2025.122671

Weichao Ding, Jiahao Liu, Wenbo Dong, Fei Luo, Chunhua Gu

{"title":"MPCMO: An improved multi-population co-evolutionary algorithm for many-objective optimization","authors":"Weichao Ding, Jiahao Liu, Wenbo Dong, Fei Luo, Chunhua Gu","doi":"10.1016/j.ins.2025.122671","DOIUrl":"10.1016/j.ins.2025.122671","url":null,"abstract":"<div><div>Many-objective optimization problems (MaOPs) are widely used in scientific research and engineering practices, which mainly consider joint optimization of multiple objectives simultaneously. Despite the numerous multi-objective evolutionary algorithms proposed in recent years, they often struggle with challenges in fitness assignment arising from objective conflicts. Meanwhile, they tend to perform well in only one aspect of convergence, diversity, and computational complexity. To address these issues, this paper proposes an improved multi-population co-evolutionary algorithm for many-objective optimization (termed MPCMO), which leverages the advantages of multi-population co-evolutionary techniques. The primary objective of MPCMO is to achieve a more balanced performance across convergence, diversity, and complexity. MPCMO comprises three essential components. Initially, an adaptive evolutionary strategy is employed to dynamically allocate evolutionary opportunities to subpopulations so as to conserve computational resources and enhance convergence. Subsequently, a migration strategy is developed to ensure a more global approximation of whole Pareto front. Additionally, an archive update-truncation strategy, based on angle selection and shift-based density estimation, is adopted to enhance diversity. We conduct comprehensive comparative experiments on a variety of many-objective benchmark problems with complicated characteristics. Experimental results demonstrate that the proposed method outperforms existing state-of-the-art algorithms in terms of both diversity and convergence.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122671"},"PeriodicalIF":6.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107071","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CF2M-Net: Cross-feature fusion and memory-constraint network for video anomaly detection CF2M-Net：视频异常检测的跨特征融合和内存约束网络

IF 6.8 1区计算机科学

Information Sciences Pub Date : 2025-09-12 DOI: 10.1016/j.ins.2025.122673

Qiming Ma , Chengyou Wang , Xiao Zhou

{"title":"CF2M-Net: Cross-feature fusion and memory-constraint network for video anomaly detection","authors":"Qiming Ma , Chengyou Wang , Xiao Zhou","doi":"10.1016/j.ins.2025.122673","DOIUrl":"10.1016/j.ins.2025.122673","url":null,"abstract":"<div><div>Video anomaly detection (VAD) aims to automatically identify anomalous events in surveillance videos that are significantly different from the normal pattern. Most existing methods learn the spatial-temporal distribution of normal features and detect deviations as anomalies. Typically, they employ autoencoders to independently learn appearance and motion features, but this separate learning limits the exploitation of their interrelation in real-world scenarios. To enhance the representation of normal patterns by capturing feature interrelation, we propose a cross-feature fusion and memory-constraint network (CF<sup>2</sup>M-Net) for VAD. Specifically, inspired by the representational ability of cross-attention in multimodal fusion, we design a cross-attention and memory-constraint (CM) module to enrich appearance features with motion information. To prevent overfitting to anomalous events, the memory-constraint module further constrains fused features within the distribution of normal patterns. We design an attention fusion (AF) decoder to predict normal features closer to the normal distribution, enhancing their separability from anomalies. By jointly modeling appearance and motion through feature fusion and memory constraints, CF<sup>2</sup>M-Net provides more discriminative normal representations for anomaly detection. Experimental evaluations on three benchmark datasets show that the CF<sup>2</sup>M-Net performs comparably with leading approaches. Moreover, the detailed evaluations indicate the effectiveness of normal representation based appearance-motion fusion features for VAD.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"723 ","pages":"Article 122673"},"PeriodicalIF":6.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107119","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0