Neurocomputing最新文献_第10页

DVC2: Deep video cascade clustering from video structures DVC2：视频结构的深度视频级联聚类

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131565

Zihua Wang , Siya Mi , Yu Zhang

{"title":"DVC2: Deep video cascade clustering from video structures","authors":"Zihua Wang , Siya Mi , Yu Zhang","doi":"10.1016/j.neucom.2025.131565","DOIUrl":"10.1016/j.neucom.2025.131565","url":null,"abstract":"<div><div>Video clustering is a critical unsupervised learning task, where category labels are unavailable, unlike in supervised video classification. The primary challenge is learning meaningful video representations without annotations to effectively group similar videos. Most existing methods extract frame-level features and apply standard clustering algorithms such as K-means, but they often fail to capture temporal relationships inherent in video data. In this paper, we introduce Deep Video Cascade Clustering (<span><math><msup><mtext>DVC</mtext><mn>2</mn></msup></math></span>), a novel unsupervised video learning paradigm. Unlike image-based clustering methods, <span><math><msup><mtext>DVC</mtext><mn>2</mn></msup></math></span> first learns an initial video representation through frame clustering, which serves as guidance, and then aligns video clustering results with both long-term and short-term structures as well as nearest neighbors. We evaluate <span><math><msup><mtext>DVC</mtext><mn>2</mn></msup></math></span> on benchmark datasets, including UCF101 and Kinetics-400, achieving state-of-the-art results. Notably, even in annotation-free scenarios where self-supervised learning with K-means already yields reasonable clustering, <span><math><msup><mtext>DVC</mtext><mn>2</mn></msup></math></span> demonstrates significantly superior performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"657 ","pages":"Article 131565"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145156963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

End-to-end transformer-based detection with density-guided query selection for small objects 基于端到端变压器的小对象密度导向查询选择检测

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131554

Nguyen Hoanh , Tran Vu Pham

{"title":"End-to-end transformer-based detection with density-guided query selection for small objects","authors":"Nguyen Hoanh , Tran Vu Pham","doi":"10.1016/j.neucom.2025.131554","DOIUrl":"10.1016/j.neucom.2025.131554","url":null,"abstract":"<div><div>Small object detection remains a persistent challenge in transformer-based detectors due to their limited localization precision and reliance on fixed query mechanisms. In this paper, we propose Hybrid Density-Transformer (HyDeTr), a novel transformer-based object detection framework designed to improve the detection of small and densely packed objects with only a slight trade-off in inference complexity. HyDeTr introduces several key innovations: (1) a Context-Selective Hybrid Attention Encoder (CS-HAE) that distills global context from low-resolution features through efficient kernelized attention while preserving local detail via deformable attention on higher-resolution maps; (2) a Density Map Prediction module that generates a spatial prior highlighting high-object-density regions, facilitating focus on crowded scenes; (3) a Density-Guided Uncertainty-Minimal Query Selection strategy that identifies the most informative query locations based on both classification confidence and predicted density, ensuring that even low-confidence small objects in dense areas are effectively queried; and (4) an improved Query Formulation with dual embeddings, consisting of a content embedding and a 4D anchor box, refined iteratively by the decoder. Our design enables precise, density-aware query initialization and scale adaptation, leading to improved recall and accuracy for small objects. Extensive evaluations demonstrate that HyDeTr outperforms existing methods in detecting small objects, offering significant accuracy gains with only a modest increase in inference complexity, thereby maintaining near real-time performance and full end-to-end trainability.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"656 ","pages":"Article 131554"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145108159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

EHGFL: Contrastive distillation for efficient heterogeneous graph few-shot learning 高效异构图少射学习的对比蒸馏

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131493

Ning Ruan , Yufei Zeng , Huan Liu , Dong Liu , Pengfei Jiao

{"title":"EHGFL: Contrastive distillation for efficient heterogeneous graph few-shot learning","authors":"Ning Ruan , Yufei Zeng , Huan Liu , Dong Liu , Pengfei Jiao","doi":"10.1016/j.neucom.2025.131493","DOIUrl":"10.1016/j.neucom.2025.131493","url":null,"abstract":"<div><div>Heterogeneous graphs (HGs), as a general modeling paradigm for multi-typed entities and complex interactions in ubiquitous real-world networks, have attracted extensive research enthusiasm. While self-supervised learning (SSL) on heterogeneous graph neural networks (HGNNs) demonstrates promising performance, existing SSL approaches face critical limitations: (1) The inherent multiple relation types and meta-path aggregations in HGNNs create prohibitive training and inference costs that restrict scalability; (2) The local message-passing paradigm confines information propagation to immediate neighborhoods, limiting the model’s ability to capture long-range dependencies and global structural patterns essential for complex reasoning tasks; (3) Task specialization necessitates costly fine-tuning of the HGNN backbones. To address these issues, we propose Efficient Heterogeneous Graph Few-shot Learning (EHGFL) to improve HGNNs’ scalability and global-structure modeling capabilities. Specifically, EHGFL first employs instance discrimination contrastive learning for self-supervised pretraining of HGNNs. To enhance efficiency, we introduce a novel cross-model contrastive distillation mechanism that transfers HGNNs’ heterogeneous structure modeling ability to a concise, globally-structure-aware multilayer perceptron. This feature-space distillation process preserves heterogeneous structure representations while avoiding expensive neighborhood aggregation and enhancing global feature awareness. Furthermore, to bridge the gap between pretraining objectives and downstream tasks, we adopt prompt tuning techniques specifically designed for the student model, enabling effective adaptation with limited labeled data. Extensive experiments on two real-world HG datasets demonstrate that the proposed EHGFL framework substantially accelerates training and inference while achieving superior few-shot node classification accuracy compared to state-of-the-art baselines.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"656 ","pages":"Article 131493"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145119825","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Generalized containment control for delayed fractional-order nonlinear multi-agent systems with unknown disturbances 具有未知扰动的延迟分数阶非线性多智能体系统的广义包容控制

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131469

Zhi Qiao , Luyang Yu , Hong Lin , Yurong Liu

{"title":"Generalized containment control for delayed fractional-order nonlinear multi-agent systems with unknown disturbances","authors":"Zhi Qiao , Luyang Yu , Hong Lin , Yurong Liu","doi":"10.1016/j.neucom.2025.131469","DOIUrl":"10.1016/j.neucom.2025.131469","url":null,"abstract":"<div><div>In this article, the issue of generalized containment control is explored for a class of delayed fractional-order nonlinear multi-agent systems (MASs) with unknown disturbances. It is assumed that the MAS under consideration has multiple dynamic leaders, and its dynamics are governed by fractional-order differential equations, and suffers from the unknown but norm-bounded external disturbances. Also, the directed graph of the MAS is assumed to have a united directed spanning tree. Furthermore, for the sake of saving communication resources, an event-triggered mechanism is introduced to regulate the signal transmission. In the presence of the external disturbances, the generalized containment control is analyzed by means of Lyapunov stability theory, algebraic graph theory, Halanay-type inequality, etc., and sufficient conditions are established to ensure that all followers ultimately enter a certain neighborhood of the convex hull formed by the leaders. In the meanwhile, it is also proven that the Zeno phenomenon can be excluded for the concerned MAS. Finally, a numerical simulation is presented to further illustrate the effectiveness of the theoretical results.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"656 ","pages":"Article 131469"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145119907","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Optimal event-triggered control for multi-agent systems with hierarchical framework 层次框架下多智能体系统的最优事件触发控制

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131477

Denghao Pang , Yechen Guo , Jinde Cao , Boxiang Li , Xiao-Wen Zhao

{"title":"Optimal event-triggered control for multi-agent systems with hierarchical framework","authors":"Denghao Pang , Yechen Guo , Jinde Cao , Boxiang Li , Xiao-Wen Zhao","doi":"10.1016/j.neucom.2025.131477","DOIUrl":"10.1016/j.neucom.2025.131477","url":null,"abstract":"<div><div>This study investigates the event-triggered optimal control problem for a class of linear second-order multi-agent systems (MASs) with external disturbances. A hierarchical framework is proposed to address the challenges that arise from the information of the coupled neighbors and external disturbances, integrating the communication, learning, and control layers. Specifically, the communication layer utilizes event-triggered mechanisms (ETMs) to transmit neighbor information, facilitating virtual consensus. The learning layer connects the communication and control layers, employing reinforcement learning (RL) to optimize tracking control with ETMs. The control layer achieves real consensus by aligning the agent states with the processed information from the communication layer. Moreover, this framework effectively mitigates the effects of coupled neighbor information on the controller and suppresses the transmission of external disturbances through the communication network. Finally, two simulation examples are used to verify the anti-interference of the hierarchical framework i.e., it’s still possible to achieve consensus after being disturbed and the effectiveness of considering the reinforcement learning layer via event-triggered mechanism which reduces the communication and learning burden to achieve optimal control.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"656 ","pages":"Article 131477"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145119909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Predicting nonlinear dynamic systems by causal physics-informed neural networks with ResNet blocks 基于ResNet块的因果物理神经网络预测非线性动态系统

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131589

Man-Hong Fan , Jun-Hao Zhao , Lin Ding , Xiao-Ying Ma , Rui-Lin Fu

{"title":"Predicting nonlinear dynamic systems by causal physics-informed neural networks with ResNet blocks","authors":"Man-Hong Fan , Jun-Hao Zhao , Lin Ding , Xiao-Ying Ma , Rui-Lin Fu","doi":"10.1016/j.neucom.2025.131589","DOIUrl":"10.1016/j.neucom.2025.131589","url":null,"abstract":"<div><div>With the continuous advancement of data computational science, the prediction of nonlinear systems has provided effective support for investigating complex problems in the field of natural sciences. Physics-Informed Neural Networks (PINNs) are playing an increasingly prominent role in nonlinear system prediction. Although PINNs have been widely applied across various engineering domains, their utilization in chaotic system prediction remains notably scarce. This paper proposes a novel causal PINNs framework integrated with ResNet blocks. On the one hand, the framework incorporates temporal weighting into the residual loss, utilizing maximum temporal weight as the training termination criterion. Additionally, an annealing strategy is adopted to adaptively adjust the causal parameters, ensuring that the model adheres to physical causality constraints throughout the training process. On the other hand, the framework employs a ResNet-block-based network, which transforms identity mappings into residual mappings. This architectural design significantly enhances training stability when utilizing deep networks. To validate the performance of the proposed method, numerical experiments are conducted on the Lorenz system, Dadras system, and Kuramoto-Sivashinsky equation. The results demonstrate that the causal PINNs with ResNet blocks significantly outperform conventional PINNs in predicting chaotic systems.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"656 ","pages":"Article 131589"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107487","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Bipartite tracking consensus for fractional-order nonlinear multiagent systems with sampled-data and input saturation 具有采样数据和输入饱和的分数阶非线性多智能体系统的二部跟踪一致性

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131472

Zhi Qiao , Luyang Yu , Yuman Li , Yurong Liu

引用次数: 0

Learning generic and specific prompts with contrastive constraints for multi-task visual scene understanding 学习通用和特定提示与多任务视觉场景理解的对比约束

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131586

Tianyu Han , Zhimin Xu , Wanying Li , Haohao Hu , Xinxin He , Song He , Peng Zan , Xiaochen Bo

{"title":"Learning generic and specific prompts with contrastive constraints for multi-task visual scene understanding","authors":"Tianyu Han , Zhimin Xu , Wanying Li , Haohao Hu , Xinxin He , Song He , Peng Zan , Xiaochen Bo","doi":"10.1016/j.neucom.2025.131586","DOIUrl":"10.1016/j.neucom.2025.131586","url":null,"abstract":"<div><div>Multi-task learning has emerged as a crucial research direction in the field of computer vision, offering improved performance and efficiency across multiple tasks. Recent studies have incorporated prompt learning into multi-task learning to enhance the interaction between prompt vectors and image representations. However, these studies fail to consider the inter-task and intra-task relations of prompt vectors under multi-task scenarios. To address this issue, we propose learning Generic and Specific Prompts (GSPrompt) with contrastive constraints for multi-task visual scene understanding. Our approach assumes that each task possesses both commonality and individuality, leading us to design two distinct types of prompt vectors: task-generic prompts and task-specific prompts. By constraining the prompt vectors through pulling task-generic prompts and pushing task-specific prompts, we enable multi-task models to learn prompts capable of adapting to multiple tasks simultaneously. Extensive experiments on NYUD-v2, PASCAL-Context, and Cityscapes show that GSPrompt learns effective prompts and achieves state-of-the-art performance. The code is publicly available at <span><span>https://github.com/teeyohan/GSPrompt-main</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"657 ","pages":"Article 131586"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145120148","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

A multi-step minimax Q-learning algorithm for two-player zero-sum Markov games 二人零和马尔可夫博弈的多步极大极小q学习算法

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131552

Shreyas S.R. , Antony Vijesh

引用次数: 0

Efficient semantic segmentation of remote sensing images through dynamic feature enhancement and multimodal alignment fusion 基于动态特征增强和多模态对齐融合的遥感图像高效语义分割

IF 6.5 2区计算机科学

Neurocomputing Pub Date : 2025-09-16 DOI: 10.1016/j.neucom.2025.131555

Wenqian Chen , Wendie Yue , Kai Chang , Hongzhi Wang , Kaijun Tan , Xinyu Liu , Xiaoyi Cao

{"title":"Efficient semantic segmentation of remote sensing images through dynamic feature enhancement and multimodal alignment fusion","authors":"Wenqian Chen , Wendie Yue , Kai Chang , Hongzhi Wang , Kaijun Tan , Xinyu Liu , Xiaoyi Cao","doi":"10.1016/j.neucom.2025.131555","DOIUrl":"10.1016/j.neucom.2025.131555","url":null,"abstract":"<div><div>Multimodal fusion has shown promising applications in integrating information from different modalities. However, existing multimodal fusion approaches in remote sensing face two main challenges: First, multimodal fusion models relying on Convolutional Neural Networks (CNNs) or Visual Transformers (ViTs) have limitations in terms of remote modeling capabilities and computational complexity, while state-space model (SSM)-based fusion models are prone to feature redundancy due to the use of multiple scanning paths, and similarly suffer from high computational complexity. Second, existing methods do not fully address inter-modal heterogeneity, leading to poor multimodal data fusion. To address these issues, we propose an efficient multimodal fusion network, AFMamba, based on the state-space model (SSM) for semantic segmentation of remote sensing images. Specifically, we design the Efficient Dynamic Visual State Space (EDVSS) module, which enhances the efficiency of the standard Mamba model by dynamically improving local features and reducing channel redundancy. Furthermore, we introduce the Cross Attention Alignment Fusion (CAAFM) module, which combines cross-image attention fusion and channel interaction alignment to effectively improve the accuracy and efficiency of cross-modal feature fusion and mitigate feature inconsistency. Experimental results demonstrate that in multimodal hyperspectral image semantic segmentation, the proposed model reduces computational complexity, measured in GFLOPs, by at least 61 % while maintaining a low parameter count, achieving optimal overall accuracy (OA) of around 92 %, and effectively balancing performance and computational efficiency.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"657 ","pages":"Article 131555"},"PeriodicalIF":6.5,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145156967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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