Neurocomputing最新文献_第5页

A positive-unlabeled learning approach for industrial anomaly detection based on self-adaptive training 基于自适应训练的工业异常检测的正无标记学习方法

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-26 DOI: 10.1016/j.neucom.2025.130488

Keyu Song, Chao Liu, Dongxiang Jiang

{"title":"A positive-unlabeled learning approach for industrial anomaly detection based on self-adaptive training","authors":"Keyu Song, Chao Liu, Dongxiang Jiang","doi":"10.1016/j.neucom.2025.130488","DOIUrl":"10.1016/j.neucom.2025.130488","url":null,"abstract":"<div><div>The availability of large amounts of data offers potential for condition monitoring and anomaly detection in complex industrial systems. However, a major obstacle arises from the prevalent issue of insufficient and sparse labeling, particularly of anomalous data. This scarcity poses a significant challenge in efficiently utilizing a large yet weakly-labeled dataset. Our work addresses this challenge by framing the problem as a highly unlabeled, imbalanced, and positive-unlabeled(PU) learning scenario, which integrates massive unlabeled data with rare anomalous data into a single training set. We introduce a self-adaptive training based approach, termed SatPU, designed for training anomaly detection models in this context. Our approach incorporates a novel pseudo labeling scheme coupled with a new dynamic re-weighting technique to facilitate self-adaptive training. These augmentations address issues related to confirmation bias and the risk of converging to trivial outcomes in the PU learning setup, thereby enhancing classification performance. Comparative experiments conducted on two industrial datasets against state-of-the-art PU learning methods validate the effectiveness of the proposed approach. The results demonstrate that SatPU exhibits superior performance when trained on industrial data that violate the select completely at random (SCAR) assumption and contain a high proportion of unlabeled samples. Furthermore, an ablation study is conducted to substantiates the necessities of the introduced augmentations for self-adaptive training.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130488"},"PeriodicalIF":5.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144190416","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

Exploiting module evolution correlation relationship for fine-grained bird image classification with structural functional representation 利用模块演化相关关系进行细粒度鸟类图像的结构函数表示分类

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-26 DOI: 10.1016/j.neucom.2025.130609

Shuang Zeng , Hai Liu , Tingting Liu , Qiuxia Liu , Minhong Wang , Bing Yang , Zhaoli Zhang

{"title":"Exploiting module evolution correlation relationship for fine-grained bird image classification with structural functional representation","authors":"Shuang Zeng , Hai Liu , Tingting Liu , Qiuxia Liu , Minhong Wang , Bing Yang , Zhaoli Zhang","doi":"10.1016/j.neucom.2025.130609","DOIUrl":"10.1016/j.neucom.2025.130609","url":null,"abstract":"<div><div>Fine-grained bird image classification (FBIC) targeting the classification of avian subspecies is facing challenges due to confusing background, partly occlusions and varied posture. In this paper, we proposed a novel module evolution correlation relationship modeling for FBIC task, which can learn structural functional representations among different functional feathers in a bird. Specially, the proposed EMECR model includes two modules, such as periodic topology mining (PTM), and multi-scale semantics alignment strategy (MSAS). The PTM module is proposed to reveal local periodic organizations with implicit functional expressions, and the MSAS is leveraged for better semantic modeling. In addition, a joint loss is designed to suppress the outliers and enforce semantic consistency. Furthermore, to better model the module evolution relationships between different functional representations and hierarchical context correlation, Mamba architecture is employed as the decoder with its linear computational complexity. Experiments on CUB-200–2011 and NABirds verify that our method can obtain robust results and significantly outperform the existing state-of-the-art FBIC methods. Extended experiments have been conducted on the Stanford Cars dataset to suggest the potential of generalizing our method on other fine-grained visual classification tasks.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130609"},"PeriodicalIF":5.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144221976","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

Synchronization control for delayed neural networks under round-robin protocol and input saturations 轮循协议和输入饱和下延迟神经网络的同步控制

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-26 DOI: 10.1016/j.neucom.2025.130474

Yonggang Chen , Kelin Zhang , Yaxue Zhao

引用次数: 0

Hierarchical cross-modal interaction network for multimodal fake news detection 多模态假新闻检测的分层跨模态交互网络

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-26 DOI: 10.1016/j.neucom.2025.130308

Zihao Li , Xiaofei Li , Jiaxin Yang, Xianghan Wang, Shuohao Li, Jun Zhang

{"title":"Hierarchical cross-modal interaction network for multimodal fake news detection","authors":"Zihao Li , Xiaofei Li , Jiaxin Yang, Xianghan Wang, Shuohao Li, Jun Zhang","doi":"10.1016/j.neucom.2025.130308","DOIUrl":"10.1016/j.neucom.2025.130308","url":null,"abstract":"<div><div>In the rapidly evolving digital information age, multimodal fake news detection has become a critical field of study to ensure the veracity of content that blends text and imagery. Existing methods utilize attention mechanism to fuse the features extracted from unimodal encoder and perform surface-level cross-modal interaction, which is inefficient due to misalignment of multimodal features and limited ability to interpret the consistency between textual and visual information. To address these issues, we present the <strong>H</strong>ierarchical <strong>C</strong>ross-<strong>M</strong>odal <strong>I</strong>nteraction <strong>N</strong>etwork (HCMIN) for multimodal fake news detection, a novel architecture that robustly achieve intra- and inter-modality interaction. Specifically, we design dual unimodal branches that independently enhance textual and visual representation, coupled with unimodal-guided co-attention mechanisms that ensure complex intra- and inter-modality feature interactions are captured. Additionally, we propose a hierarchical mutual contrastive learning strategy, which refines the alignment and interactions of modalities at both feature and decision levels, utilizing contrastive loss to spotlight discriminative characteristics of fake news. Extensive experiments on three benchmarks demonstrate the superiority of our HCMIN.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130308"},"PeriodicalIF":5.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144190423","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 zeroing neural dynamics algorithm for discrete-time nonlinear optimization with linear equality constraint and perturbation inhibition 具有线性等式约束和扰动抑制的离散时间非线性优化归零神经动力学算法

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-26 DOI: 10.1016/j.neucom.2025.130492

Chao Jiang , Jiacheng Wang , Ruxin Zhao , Dimitrios K. Gerontitis , Yang Shi

{"title":"A zeroing neural dynamics algorithm for discrete-time nonlinear optimization with linear equality constraint and perturbation inhibition","authors":"Chao Jiang , Jiacheng Wang , Ruxin Zhao , Dimitrios K. Gerontitis , Yang Shi","doi":"10.1016/j.neucom.2025.130492","DOIUrl":"10.1016/j.neucom.2025.130492","url":null,"abstract":"<div><div>As neural network technology gallops forward, more and more applied mathematical problems can be solved by such an intelligent algorithm. The traditional zeroing neural dynamics (ZND) can only solve problems effectively without perturbation. However, in the practical environment, the existence of perturbation may seriously affect the computational performance of intelligent algorithm. This paper investigates the discrete time-varying nonlinear optimization problem with linear equality constraints in the perturbation environment. With the help of integration technique, the model has the excellent performance of perturbation inhibition. First of all, the Lagrange multiplier method is used to construct a continuous-time ZND algorithm. Secondly, the widely-used discretization formulas (i.e., Euler-type discretization formula, Taylor-type discretization formula and five-instant-type discretization formula) are used to obtain different discrete-time ZND (DT-ZND) algorithms. Finally, in the experiments, compared with traditional DT-ZND algorithms, the proposed DT-ZND algorithms (i.e., Eu-DT-ZND algorithm, Ta-DT-ZND algorithm and FI-DT-ZND algorithm) have high accuracy and excellent performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130492"},"PeriodicalIF":5.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144195640","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

Neural network-based adaptive prescribed-time bipartite flocking for uncertain networked multi-agent systems 基于神经网络的不确定网络多智能体系统自适应定时二部蜂拥

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-25 DOI: 10.1016/j.neucom.2025.130452

Xian Qing, Weihao Li, Bowen Chen, Boxian Lin, Mengji Shi, Kaiyu Qin

{"title":"Neural network-based adaptive prescribed-time bipartite flocking for uncertain networked multi-agent systems","authors":"Xian Qing, Weihao Li, Bowen Chen, Boxian Lin, Mengji Shi, Kaiyu Qin","doi":"10.1016/j.neucom.2025.130452","DOIUrl":"10.1016/j.neucom.2025.130452","url":null,"abstract":"<div><div>Flocking is a fundamental self-organizing behavior observed in networked agent systems (NASs), wherein agents achieve coordinated group dynamics through mutual interactions. While in dynamic environmental contexts, the demand for flocking behavior to demonstrate rapid responsiveness and robust stability becomes more critical. With this in mind, this paper addresses the adaptive bipartite flocking control problem for NASs, particularly in the presence of compound uncertainties and convergence time constraints. A robust adaptive neural network-based prescribed-time bipartite flocking controller is developed to ensure that, despite uncertainties, all agents achieve flocking behavior within a predefined time. Notably, the settling time can be predefined and remains independent of system parameters such as controller gain, initial agent states, and the communication topology among agents. Additionally, by analyzing the stability conditions of the closed-loop error system, an adaptive weight update law for the neural network estimator is formulated. This updated law allows for effective uncertainty estimation through backpropagation of the flocking control error. Finally, the effectiveness and superiority of the proposed prescribed-time bipartite flocking control scheme are validated through numerical simulations.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"646 ","pages":"Article 130452"},"PeriodicalIF":5.5,"publicationDate":"2025-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144168959","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

Adaptive continuous fractional-order nonsingular terminal sliding mode control based on neural network for PMLM system with actuator saturation 基于神经网络的连续分数阶非奇异终端滑模控制

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-24 DOI: 10.1016/j.neucom.2025.130468

Haoran Zhang , Jinyong Yu , Pengwei Shi , Shenglin Hu , Lin Zhao

{"title":"Adaptive continuous fractional-order nonsingular terminal sliding mode control based on neural network for PMLM system with actuator saturation","authors":"Haoran Zhang , Jinyong Yu , Pengwei Shi , Shenglin Hu , Lin Zhao","doi":"10.1016/j.neucom.2025.130468","DOIUrl":"10.1016/j.neucom.2025.130468","url":null,"abstract":"<div><div>This paper addresses the trajectory tracking control problem for permanent magnet linear motor (PMLM) systems, presenting a novel fractional-order nonsingular terminal sliding mode control (SMC) strategy integrated with a radial basis function (RBF) neural network approximator. The proposed method leverages a fractional-order sliding surface to achieve superior convergence precision compared to conventional nonsingular SMC approaches in finite time. To handle unknown disturbances and lumped uncertainties, an RBF neural network is employed for adaptive approximation, complemented by a dynamically adjustable compensation term to mitigate approximation errors under varying conditions. Actuator saturation, a practical constraint, is explicitly addressed through an auxiliary system design, ensuring robust motion performance. Compared with integer-order sliding mode control, the proposed scheme demonstrates faster convergence and enhanced tracking accuracy, validated theoretically via Lyapunov stability analysis. Simulation results are provided to confirm the efficacy of the control strategy against state-of-the-art methods, with root mean square error (RMSE) serving as a quantitative metric to support the experimental findings. This work has made obvious advances by integrating fractional-order dynamics, adaptive neural approximation, and saturation compensation, offering a robust solution with clear performance improvements for PMLM system.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"646 ","pages":"Article 130468"},"PeriodicalIF":5.5,"publicationDate":"2025-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144168960","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

Updating KernelCanvas for weightless graph classification 更新KernelCanvas用于无权重图分类

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-24 DOI: 10.1016/j.neucom.2025.130458

Raul Bezerra Barbosa , Diego Leonel Cadette Dutra , Priscila Machado Vieira Lima , Diego Carvalho , Felipe Maia Galvão França

引用次数: 0

Release probability distribution modulates attractor state stability and mobility in continuous-attractor neural networks 释放概率分布调节连续吸引子神经网络中吸引子状态的稳定性和迁移性

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-24 DOI: 10.1016/j.neucom.2025.130450

Zihan Liu , Flavia Nathaline Chanentia , Patteera Supvithayanond , Chi Chung Alan Fung

引用次数: 0

A generalized classification of human sperm head morphology via Contrastive Meta-learning with Auxiliary Tasks 通过辅助任务对比元学习对人类精子头部形态的广义分类

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-05-24 DOI: 10.1016/j.neucom.2025.130454

Yuh-Shyan Chen, Yu-Chi Chang, Jia-Yi Hong

{"title":"A generalized classification of human sperm head morphology via Contrastive Meta-learning with Auxiliary Tasks","authors":"Yuh-Shyan Chen, Yu-Chi Chang, Jia-Yi Hong","doi":"10.1016/j.neucom.2025.130454","DOIUrl":"10.1016/j.neucom.2025.130454","url":null,"abstract":"<div><div>Semen analysis is the primary method for evaluating male infertility, with sperm morphology being a critical indicator. However, existing methods for classifying human sperm head morphology (HSHM) often lack cross-domain generalizability. To address this limitation, we propose an enhanced meta-learning algorithm that learns invariant features across tasks, improving generalization by transferring knowledge to new tasks. To mitigate gradient conflicts in multi-task learning, we separate meta-training tasks into primary and auxiliary tasks. This approach, in conjunction with auxiliary tasks, enhances the model’s generalization using diverse HSHM datasets. We introduce the Contrastive Meta-Learning with Auxiliary Tasks (HSHM-CMA) algorithm, which integrates localized contrastive learning in the outer loop of meta-learning to exploit invariant sperm morphology features across domains. This improves task convergence and adaptation to new categories. In our evaluation, we assess the model’s generalization performance using three testing objectives: the same dataset with different HSHM categories, different datasets with the same HSHM categories, and different datasets with different HSHM categories. Our evaluation across these objectives demonstrates that HSHM-CMA outperforms existing meta-learning approaches, achieving accuracies of 65.83%, 81.42%, and 60.13%, respectively.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"645 ","pages":"Article 130454"},"PeriodicalIF":5.5,"publicationDate":"2025-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144135136","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