Neurocomputing最新文献_第3页

Explaining Graph Neural Networks with mixed-integer programming

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-19 DOI: 10.1016/j.neucom.2025.130214

Blake B. Gaines , Chunjiang Zhu , Jinbo Bi

{"title":"Explaining Graph Neural Networks with mixed-integer programming","authors":"Blake B. Gaines , Chunjiang Zhu , Jinbo Bi","doi":"10.1016/j.neucom.2025.130214","DOIUrl":"10.1016/j.neucom.2025.130214","url":null,"abstract":"<div><div>Graph Neural Networks (GNNs) provide state-of-the-art graph learning performance, but their lack of transparency hinders our ability to understand and trust them, ultimately limiting the areas where they can be applied. Many methods exist to explain individual predictions made by GNNs, but there are fewer ways to gain more general insight into the patterns they have been trained to identify. Most existing methods for model-level GNN explanations attempt to generate graphs that exemplify these patterns, but the discreteness of graphs and the nonlinearity of deep GNNs make finding such graphs difficult. In this paper, we formulate the search for an explanatory graph as a mixed-integer programming (MIP) problem, in which decision variables specify the explanation graph and the objective function represents the quality of the graph as an explanation for a GNN’s predictions of an entire class in the dataset. This approach, which we call MIPExplainer, allows us to directly optimize over the discrete input space and find globally optimal solutions with a minimal number of hyperparameters. MIPExplainer outperforms existing methods in finding accurate and stable explanations on both synthetic and real-world datasets. Code is available at <span><span>https://github.com/blake-gaines/MIPExplainer</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130214"},"PeriodicalIF":5.5,"publicationDate":"2025-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876660","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 spiking convolutional network for decoding motor imagery patterns

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.130303

Zimeng Zhu , Fei Liu , Siqi Mu , Kuan Tao

{"title":"Adaptive spiking convolutional network for decoding motor imagery patterns","authors":"Zimeng Zhu , Fei Liu , Siqi Mu , Kuan Tao","doi":"10.1016/j.neucom.2025.130303","DOIUrl":"10.1016/j.neucom.2025.130303","url":null,"abstract":"<div><div>Motor imagery (MI) decoding is crucial for the advancement of brain-computer interface (BCI) technologies. However, existing models often suffer from susceptibility to noise and lack biological interpretability. In this study, we introduce the Adaptive Firing Threshold Spiking Convolutional Network (AFTSC-Net), which enhances the biological relevance of spiking neural networks (SNNs) by integrating an adaptive firing rate mechanism with the spatial feature extraction capabilities of convolutional neural networks (CNNs). Additionally, we refined surrogate gradient functions through enhanced spiking neuron mechanisms, significantly reducing computational power consumption while improving the accuracy of MI pattern recognition from electroencephalography (EEG) signals. To validate the efficacy of AFTSC-Net, we conducted experiments with 36 elite athletes from soccer, basketball, and table tennis, performing a comprehensive analysis of neural activity across various motor imagery tasks. The model not only demonstrated superior performance on the athlete dataset but also achieved the state-of-the-art results on public benchmark datasets, surpassing existing methods in terms of accuracy and computational efficiency. These findings highlight the potential of biologically inspired neural networks to enhance MI decoding accuracy and robustness, setting a new standard for real-time BCI applications.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130303"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143844491","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

Overcoming the Price of Anarchy by Steering with Recommendations

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.129993

Cesare Carissimo, Marcin Korecki, Damian Dailisan

{"title":"Overcoming the Price of Anarchy by Steering with Recommendations","authors":"Cesare Carissimo, Marcin Korecki, Damian Dailisan","doi":"10.1016/j.neucom.2025.129993","DOIUrl":"10.1016/j.neucom.2025.129993","url":null,"abstract":"<div><div>Varied real world systems such as transportation networks, supply chains and energy grids present coordination problems where many agents must learn to share resources. It is well known that the independent and selfish interactions of agents in these systems may lead to inefficiencies, often referred to as the ‘Price of Anarchy’. Effective interventions that reduce the Price of Anarchy while preserving individual autonomy are of great interest. In this paper we explore recommender systems as one such intervention mechanism. We start with the Braess Paradox, a congestion game model of a routing problem related to traffic on roads, packets on the internet, and electricity on power grids. Following recent literature, we model the interactions of agents as a repeated game between <span><math><mi>Q</mi></math></span>-learners, a common type of reinforcement learning agents. This work introduces the Learning Dynamic Manipulation Problem, where an external recommender system can strategically trigger behavior by picking the states observed by <span><math><mi>Q</mi></math></span>-learners during learning. Our computational contribution demonstrates that appropriately chosen recommendations can robustly steer the system towards convergence to the social optimum, even for many players. Our theoretical and empirical results highlight that increases in the recommendation space can increase the steering potential of a recommender system, which should be considered in the design of recommender systems.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 129993"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870069","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

Exploring inter- and intra-modal relations in compositional zero-shot learning

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.130213

Xiao Zhang, Hui Chen, Haodong Jing, Yongqiang Ma, Nanning Zheng

{"title":"Exploring inter- and intra-modal relations in compositional zero-shot learning","authors":"Xiao Zhang, Hui Chen, Haodong Jing, Yongqiang Ma, Nanning Zheng","doi":"10.1016/j.neucom.2025.130213","DOIUrl":"10.1016/j.neucom.2025.130213","url":null,"abstract":"<div><div>Compositional Zero-Shot Learning (CZSL) aims to recognize unknown compositions by leveraging learned concepts of states and objects. Prior methods have typically emphasized either inter-modal relation for multi-modal fusion, ignoring the entanglement within state–object pairs, or solely intra-modal relation for enhancing representations, neglecting the association between vision and language domains. To tackle these limitations, we propose a CZSL framework that simultaneously learns inter- and intra-modal relations to improve image-label alignment. Firstly, we explore <strong>inter-modal relation</strong> to enable image features to grasp the cross-modal information from states and objects. The image–text fusion method facilitates the modeling of text-aware image features and image-aware text features, improving the model’s compositional recognition capability. Secondly, due to the contextuality within state–object pairs, we further explore <strong>intra-modal relation</strong> to exploit semantic information from various representation subspaces, facilitating the comprehensive semantic expression of text features. Moreover, we propose a <strong>composition fusion module</strong> to establish semantic entanglement within state–object compositions. Extensive experiments demonstrate that our method significantly surpasses the state-of-the-art methods in both closed-world and open-world settings.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130213"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143859846","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

Distributed event-triggered state estimation for renewable microgrids subject to incomplete observations

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.130220

Peixia Gao , Wen Chen , Chaoqing Jia , Jiawen Zhang , Hongxu Zhang , Jun Hu

{"title":"Distributed event-triggered state estimation for renewable microgrids subject to incomplete observations","authors":"Peixia Gao , Wen Chen , Chaoqing Jia , Jiawen Zhang , Hongxu Zhang , Jun Hu","doi":"10.1016/j.neucom.2025.130220","DOIUrl":"10.1016/j.neucom.2025.130220","url":null,"abstract":"<div><div>In this paper, we investigate the distributed state estimation problem for renewable microgrids (RMGs) with incomplete observations, where information transmission is governed by the event-triggered communication criterion. The missing measurements with the description of uncertain occurrence probabilities (UOPs) are considered and modeled via the integration of nominal probabilities and the associated bounds. In addition, an event-triggered mechanism involving some parameters is employed to improve reliability of communication by transmitting measurements under specific triggered conditions. The aim of this paper is to design a distributed event-triggered state estimation algorithm against missing measurements under UOPs that guarantees the existence of an upper bound on the estimation error covariance (EEC) with satisfactory algorithm performance. Afterwards, the gain matrix of the corresponding state estimator is properly designed by minimizing the trace of the upper bound on the EEC. Besides, the boundedness of the upper bound of EEC is further ensured by providing a sufficient condition. Subsequently, we discuss the monotonicity relationship with respect to the trace of upper bound and the nominal occurrence probability of missing measurements. Finally, a simulation experiment with comparisons is conducted on RMGs with two distributed generation units to demonstrate the effectiveness of newly designed state estimation algorithm.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130220"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143844493","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

Incremental learning tracking control of complex dynamical trajectories for robotic manipulator

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.130215

Hainan Yang, Tao Zhao, Jia-Yu Zhao, Jianjian Zhao, Peng Qin

{"title":"Incremental learning tracking control of complex dynamical trajectories for robotic manipulator","authors":"Hainan Yang, Tao Zhao, Jia-Yu Zhao, Jianjian Zhao, Peng Qin","doi":"10.1016/j.neucom.2025.130215","DOIUrl":"10.1016/j.neucom.2025.130215","url":null,"abstract":"<div><div>The inherent complexity and unpredictability of complex trajectories pose significant challenges for tracking them using a robotic manipulator’s end effector. To address this challenge, a novel incremental learning control scheme is proposed and applied to the tracking of complex dynamic trajectories by the end effector (TCDTEE). Stochastic input torque signals are exerted on the end effector of robotic manipulator so that the modeling data are collected. Then, the T-S fuzzy model (also known as an inverse model) is constructed for TCDTEE via an ensemble fuzzy framework, making it possible to obtain a control signal that can accomplish most control tasks under nominal conditions. However, the controller with fixed structure suffers from the accumulation of tracking errors and may be incapable of addressing some unexpected dynamic changes. To overcome such two problems, an interval type-2 evolving fuzzy neural network (IT2EFNN) is designed to estimate the ideal torque control rates. The IT2EFNN initializes its structure from scratch and optimizes it through the elimination of redundant rules based on the quality assessment of the fuzzy rules, thereby enhancing tracking control accuracy and system responsiveness. Meanwhile, with the help of Lyapunov analysis approach, the stability of the IT2EFNN is guaranteed. Finally, the effectiveness of the proposed method has been validated through simulations and experiments.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130215"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143859841","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 semi-supervised multi-task assisted method for ultrasound medical image segmentation

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-17 DOI: 10.1016/j.neucom.2025.130217

Honghe Li, Jinzhu Yang, Mingjun Qu, Yong Feng

{"title":"A semi-supervised multi-task assisted method for ultrasound medical image segmentation","authors":"Honghe Li, Jinzhu Yang, Mingjun Qu, Yong Feng","doi":"10.1016/j.neucom.2025.130217","DOIUrl":"10.1016/j.neucom.2025.130217","url":null,"abstract":"<div><div>The accurate segmentation of the left ventricle in echocardiography is critical for assessing cardiac function, but challenges such as blurred boundaries, high morphological variability, and limited annotated data hinder the performance of traditional methods. Recent advances in semi-supervised learning show promise in leveraging unlabeled data, yet existing techniques often rely on network or data-level perturbation, which may not fully exploit spatial and positional information essential for precise segmentation. To address these challenges, we propose a novel multi-task assisted semi-supervised segmentation framework. Our method combines segmentation, landmark detection, and image reconstruction into a unified model with a shared encoder and dual decoders. A feature cross-fusion module based on a cross-attention mechanism integrates features across tasks to enhance spatial and positional awareness. We have also introduced a contrastive learning mechanism to refine the segmentation boundaries, especially in areas where the edges are blurred. In addition, we utilize multi-scale supervision to better adapt the model to targets of different scales. The framework employs an exponential moving average student–teacher model to effectively utilize unlabeled data for training. Experiments on CAMUS and EchoNet-Dynamic datasets demonstrate that the proposed method achieves state-of-the-art performance, delivering near fully-supervised results with only 10%–20% labeled data. This highlights its potential for high-quality segmentation in low-annotation scenarios, outperforming existing semi-supervised learning methods.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130217"},"PeriodicalIF":5.5,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870071","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

An improved predictive function control algorithm via wavelet neural network for urban rail train tracking control

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-16 DOI: 10.1016/j.neucom.2025.130223

Longda Wang , Gang Liu , Chuanfang Xu

{"title":"An improved predictive function control algorithm via wavelet neural network for urban rail train tracking control","authors":"Longda Wang , Gang Liu , Chuanfang Xu","doi":"10.1016/j.neucom.2025.130223","DOIUrl":"10.1016/j.neucom.2025.130223","url":null,"abstract":"<div><div>This study proposes a novel, effective improved predictive function control based on a wavelet neural network (IPFC-WNN) for urban rail train tracking control. Specifically, the step function and Morlet wavelet function were chosen as the base function together, and an adaptive nonlinear online adjustment function of the softening factor was proposed based on the fuzzy satisfaction of system performance and optimisation factor. The maximum and minimum softening factors for a simple straight line can also be set appropriately by a wavelet neural network according to the actual situation. To effectively improve the control performance of the predictive function control algorithm for urban rail train tracking, calculation of additional resistance with a multiparticle model was adopted, and parameters for the adaptive nonlinear online softening factor adjustment function were set using a wavelet neural network to improve the comprehensive performance quality for urban rail train tracking control. Considering the scenario of urban rail train tracking control from Bayi Road to Yongan Four Seasons, which is located in the second-phase project of Dalian Urban Rail Transit Line 13, as the hardware-in-the-loop test object, the proposed IPFC-WNN and three improved control algorithms were used for comparative verification. The test results showed that the proposed IPFC-WNN can significantly improve the performance of the control system, and quality indicators such as energy saving, precise parking, punctuality, and comfort of the system were significantly improved. Hence, the good tracking control for train operation using the proposed IPFC-WNN was verified.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130223"},"PeriodicalIF":5.5,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143850730","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

Tensor networks for explainable machine learning in cybersecurity

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-16 DOI: 10.1016/j.neucom.2025.130211

Borja Aizpurua , Samuel Palmer , Román Orús

引用次数: 0

Rethinking attention mechanism for enhanced pedestrian attribute recognition

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-04-16 DOI: 10.1016/j.neucom.2025.130236

Junyi Wu , Yan Huang , Min Gao , Yuzhen Niu , Yuzhong Chen , Qiang Wu

{"title":"Rethinking attention mechanism for enhanced pedestrian attribute recognition","authors":"Junyi Wu , Yan Huang , Min Gao , Yuzhen Niu , Yuzhong Chen , Qiang Wu","doi":"10.1016/j.neucom.2025.130236","DOIUrl":"10.1016/j.neucom.2025.130236","url":null,"abstract":"<div><div>Pedestrian Attribute Recognition (PAR) plays a crucial role in various computer vision applications, demanding precise and reliable identification of attributes from pedestrian images. Traditional PAR methods, though effective in leveraging attention mechanisms, often suffer from the lack of direct supervision on attention, leading to potential overfitting and misallocation. This paper introduces a novel and model-agnostic approach, Attention-Aware Regularization (AAR), which rethinks the attention mechanism by integrating causal reasoning to provide direct supervision of attention maps. AAR employs perturbation techniques and a unique optimization objective to assess and refine attention quality, encouraging the model to prioritize attribute-specific regions. Our method demonstrates significant improvement in PAR performance by mitigating the effects of incorrect attention and fostering a more effective attention mechanism. Experiments on standard datasets showcase the superiority of our approach over existing methods, setting a new benchmark for attention-driven PAR models.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130236"},"PeriodicalIF":5.5,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870073","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