Neurocomputing最新文献

{"title":"Feature-level attention network with group-aware interest modeling for sequential recommendation","authors":"Wei Jiang ,&nbsp;Yongquan Fan ,&nbsp;Yajun Du ,&nbsp;Xianyong Li ,&nbsp;Xiaomin Wang","doi":"10.1016/j.neucom.2025.130302","DOIUrl":"10.1016/j.neucom.2025.130302","url":null,"abstract":"<div><div>Sequential recommendation focuses on modeling user preferences based on their historical interaction sequences to predict future behaviors with greater precision. Incorporating feature-level information beyond item IDs has become a crucial approach to improving the performance of the recommendation system. However, existing methods overlook the hierarchical group relationships among users. This limitation prevents these methods from fully capturing user preferences, leading to an incomplete understanding of their true interests. Meanwhile, effectively leveraging multi-source information in recommendation systems remains a significant challenge. Existing methods typically rely on simple techniques such as pooling or concatenation to integrate information from different sources, which could degrade overall performance. To address these limitations, we propose a novel approach: <u><strong>F</strong></u>eature-level <u><strong>A</strong></u>ttention <u><strong>N</strong></u>etwork with <u><strong>G</strong></u>roup-aware <u><strong>I</strong></u>nterest <u><strong>M</strong></u>odeling for Sequential Recommendation (FANGIM). Specifically, we first employ two distinct encoders to generate user embeddings at different level. Next, we introduce a group clustering module, which identifies potential interest groups at multiple granularities and derives user group interest embeddings for both item and feature level interactions. Furthermore, we design a multi-source representation fusion module that effectively integrates information from diverse sources, reducing the semantic gap between different representation spaces. Additionally, we incorporate contrastive learning within this module to ensure consistency between the different levels of representations. Finally, extensive experiments demonstrate that FANGIM outperforms state-of-the-art baselines across four datasets.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130302"},"PeriodicalIF":5.5,"publicationDate":"2025-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876664","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}

{"title":"Fast blind image deblurring via patch-wise maximum content-weighted prior","authors":"Zheng Guo ,&nbsp;Wei Yan ,&nbsp;Zirui Zhang,&nbsp;Zhixiang Wu,&nbsp;Zhenhua Xu,&nbsp;Chunyong Wang,&nbsp;Jiancheng Lai,&nbsp;Zhenhua Li","doi":"10.1016/j.neucom.2025.130267","DOIUrl":"10.1016/j.neucom.2025.130267","url":null,"abstract":"<div><div>Blind image deblurring aims to derive the kernel and corresponding clear version solely from blurred images. This paper introduces an innovative blind image deblurring method based on the patch-wise maximum content-weighted prior (<em>PMCW</em>). Our work originates from the intuitive observation that the maximum content-weighted value of non-overlapping patches will significantly decrease after blurring degradation, which we demonstrate both mathematically and empirically. Building upon this observation, we propose a novel blind deblurring model combining <span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span>-regularized <em>PMCW</em> and <span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span>-regularized gradient prior, and develop an efficient solution algorithm utilizing projected alternating minimization (PAM). Qualitative and quantitative evaluation results on multiple benchmark datasets indicate that our proposed model achieves optimal performance, surpassing state-of-the-art algorithms in solving efficiency and various quantitative metrics.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130267"},"PeriodicalIF":5.5,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876662","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}

{"title":"A fast solution method for the Dynamic Flexible Pickup and Delivery Problem with task allocation fairness for multiple vehicles","authors":"Zhihui Sun,&nbsp;Ran Tian,&nbsp;Jiarui Wu,&nbsp;Xin Lu,&nbsp;Jinshi Wang","doi":"10.1016/j.neucom.2025.130266","DOIUrl":"10.1016/j.neucom.2025.130266","url":null,"abstract":"<div><div>The Dynamic Flexible Pickup and Delivery Problem (DFPDP) originates from the actual needs of multi-warehouse management strategies and is one of the important challenges currently facing the field of logistics and distribution. In DFPDP, it is necessary to address dynamic order fluctuations, quickly plan heterogeneous fleet routes, ensure fairness in task allocation, and minimize total travel time under time window constraints. However, there is currently little research on this issue, and traditional heuristic algorithms make it difficult to quickly find a solution to this problem. First, we propose a Multimodal Constraint Dynamic Scheduling Mechanism (MCDSM) to select a vehicle with the lowest current time consumption to make task allocation between vehicles as fair as possible. Second, we propose a Parallel Encoder-Serial Decoder model integrating Variable-length Sequences (PESDVS), in which the variable-length sequences designed can effectively handle the generation of dynamic orders and the changes in the number of pickup and delivery locations, while the trained model can adapt itself to different order scenarios. In addition, the model improves the quality of order decisions through a parallel encoder and serial decoder structure to minimize the total traveling time of the fleet. Extensive experimental results demonstrate that our method has excellent performance and good generalization ability under different order sizes. At the same time, compared with heuristic algorithms, our method can quickly find a feasible solution to the problem and the task allocation between vehicles is relatively fair.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130266"},"PeriodicalIF":5.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874515","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}

{"title":"Pick and mix reliable pseudo labels for scribble-supervised medical image segmentation","authors":"Jiawei Su ,&nbsp;Zhiming Luo ,&nbsp;Dazhen Lin ,&nbsp;Lihui Lin ,&nbsp;Shaozi Li","doi":"10.1016/j.neucom.2025.130293","DOIUrl":"10.1016/j.neucom.2025.130293","url":null,"abstract":"<div><div>Scribble-supervised segmentation methods have attracted significant attention in the field of medical imaging because of their potential to alleviate the data annotation burden. However, these methods often underperform due to a lack of sufficient supervision. Various methods have attempted to enrich the supervisory signals in different ways, including mixing pseudo labels from different samples (referred as Mixup-based method). However, these methods primarily focus on the quantity of enriched supervisory signals, disregarding their quality. This oversight presents a major drawback in that low-quality signals are often contaminated with the noise, thus can lead to undermine performance. Therefore, it is crucial to not only introduce diverse supervisory signals but also ensure their quality and reliability. Motivated by this understanding, we propose a new framework named Pick &amp; Mix, which builds upon the Mixup-based method. In the first step, we leverage the consistency of intra-class features to assess the reliability of pseudo-labels. To enhance the quality of pseudo labels, we assign lower weights to those unreliable pseudo-labels to mitigate the noise effect in the training process. Furthermore, we utilize a threshold to pick reliable pseudo-labels based on their reliability score. In the second step, we mix the reliable pseudo-labels from various samples and generate corresponding mixed images to provide richer supervisory signals for model training. In this manner, we enhance the quality of supervisory signals by generating and picking reliable ones, as well as enrich the quantity of these signals through a process of mixing. Finally, we evaluated our framework on three publicly available datasets: ACDC, MSCMRseg, and BraTS2020. The experimental results demonstrate that our approach achieves state-of-the-art performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130293"},"PeriodicalIF":5.5,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876663","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}

{"title":"E-Mamba: An efficient Mamba point cloud analysis method with enhanced feature representation","authors":"Dengao Li ,&nbsp;Zhichao Gao ,&nbsp;Shufeng Hao ,&nbsp;Ziyou Xun ,&nbsp;Jiajian Song ,&nbsp;Jie Cheng ,&nbsp;Jumin Zhao","doi":"10.1016/j.neucom.2025.130201","DOIUrl":"10.1016/j.neucom.2025.130201","url":null,"abstract":"<div><div>As a key technology for three-dimensional space analysis, point cloud analysis is widely used in many fields such as automated machinery, unmanned vehicles and virtual reality. Learning local and global features of point cloud is crucial for gaining a deep understanding of point cloud data. In point cloud local feature learning, sub-clouds with center coordinates subtracted are usually used as point patches, which are then input into mini-PointNet to enhance the point cloud feature representation. However, this method has a high dependence on the point cloud density, which affects the model performance. In this work, we introduce E-Mamba, a new model for efficient point cloud analysis. We use Scalable Embedding to rescale and patch embedding sub-clouds, which improves the model’s feature representation and generalization capabilities for point cloud. In addition, we also introduced Holosync Reordering Pooling to reorder tokens while preserving the original sequence, and used the hybrid pooling method to extract global features. In this way, the model fully utilizes the periodicity of Mamba while achieving good generalization and global feature extraction capabilities. We conduct extensive experiments on ModelNet40, ScanObjectNN, and ShapeNetPart datasets. The results show that E-Mamba can achieve superior performance while significantly reducing GPU memory usage and FLOPs, whether pre-trained or not.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130201"},"PeriodicalIF":5.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870072","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}

{"title":"An introspection of graph structure learning: A graph skeleton extraction via minimum dominating set","authors":"Zifeng Ye,&nbsp;Aifu Han,&nbsp;Guolin Chen,&nbsp;Xiaoxia Huang","doi":"10.1016/j.neucom.2025.130257","DOIUrl":"10.1016/j.neucom.2025.130257","url":null,"abstract":"<div><div>Graph structure learning (GSL) is a data-driven learning approach that has garnered widespread attention in recent years. Nevertheless, the insufficient understanding of latent graph properties poses various challenges for effective graph modeling. This raises the following question: What type of graph skeleton can preserve the most crucial latent properties that significantly impact the performance of graph neural networks (GNNs) in downstream tasks? To this end, we have conducted a comprehensive study on three key graph properties: homophily, degree distribution, and connected components, and determined how these factors influence semi-supervised node classification tasks. Specifically, the influence of homophily on GNN performance is rigorously assessed. Motivated by the analysis, a dual-sparsity graph extraction method, based on the minimum dominating set (MDS), is proposed, to intelligently select informative edges under a given edge sampling ratio. This method effectively captures the scale-free characteristics of the degree distribution, and prioritizes the preservation of node connectivity. Experimental results show that homophily is a key factor in achieving high GNN accuracy. Additionally, the degree distribution and connected components describe the connectivity patterns of the graph from both local and global topological perspectives, which are highly correlated with node classification performance under the GNN message-passing mechanism. This work reveals the necessity of considering the graph skeleton and provides a stepping stone for facilitating GSL using these latent graph properties.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130257"},"PeriodicalIF":5.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876661","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}

{"title":"SSSI-L2p: An EEG extended source imaging algorithm based on the structured sparse regularization with L2p-Norm","authors":"Shu Peng ,&nbsp;Hongyu Li ,&nbsp;Yujie Deng ,&nbsp;Hong Yu ,&nbsp;Weibo Yi ,&nbsp;Ke Liu","doi":"10.1016/j.neucom.2025.130250","DOIUrl":"10.1016/j.neucom.2025.130250","url":null,"abstract":"<div><div>Electroencephalographic (EEG) source imaging (ESI) aims to estimate brain activity locations and extents. ESI is crucial for studying brain functions and detecting epileptic foci. However, accurately reconstructing extended sources remains challenging due to high susceptibility of EEG signals to interference and the underdetermined nature of the ESI problem. In this study, we introduce a new ESI algorithm, Structured Sparse Source Imaging based on <span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span>-norm (SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span>), to estimate potential brain activities. SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span> utilizes the mixed <span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span>-norm (<span><math><mrow><mn>0</mn><mo>&lt;</mo><mi>p</mi><mo>&lt;</mo><mn>1</mn></mrow></math></span>) to enforce spatial–temporal constraints within a structured sparsity regularization framework. By leveraging the alternating direction method of multipliers (ADMM) and iteratively reweighted least squares (IRLS) algorithm, the challenging optimization problem of SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span> can be effectively solved. We showcase the superior performance of SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span> over benchmark ESI methods through numerical simulations and human clinical data. Our results demonstrate that sources reconstructed by SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span> exhibit high spatial resolution and clear boundaries, highlighting its potential as a robust and effective ESI technique. Additionally, we have shared the source code of SSSI-<span><math><msub><mrow><mi>L</mi></mrow><mrow><mn>2</mn><mi>p</mi></mrow></msub></math></span> at <span><span>https://github.com/Mashirops/SSSI-L2p.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130250"},"PeriodicalIF":5.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870070","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}

{"title":"Kernel broad learning cauchy conjugate gradient algorithm for online chaotic time series prediction","authors":"Liyun Su,&nbsp;Xiaoyi Wang","doi":"10.1016/j.neucom.2025.130234","DOIUrl":"10.1016/j.neucom.2025.130234","url":null,"abstract":"<div><div>Accurate prediction of nonlinear systems in non-Gaussian noise environments has long been a significant challenge in the fields of statistical data analysis and time series modeling. To address this issue, this paper proposes an improved Cauchy Conjugate Gradient algorithm based on a kernel broad learning feature extraction strategy (Kernel Broad Learning Cauchy Conjugate Gradient, KBLCCG). This algorithm integrates kernel mapping with broad learning systems, forming a dual feature extraction mechanism that effectively captures the complex nonlinear structures of chaotic time series while preserving their inherent dynamic chaotic characteristics. The KBLCCG algorithm utilizes its robust feature extraction capabilities through the dual extraction mechanism of kernel mapping and broad learning systems, effectively capturing the intricate nonlinear structures present in time series data. The kernel broad learning strategy mitigates the phenomenon of kernel matrix size expansion during the iterative process, thereby reducing the computational burden and enhancing the algorithm's robustness. The Cauchy Conjugate Gradient method is employed to optimize the reduced-dimensional feature data, efficiently addressing the nonlinear prediction problem of the target sequence. Empirical analysis using simulation data and actual financial data (including the Lorenz system, Shanghai Composite Index, and CSI 300 Index) validates the performance of this method. Experimental results indicate that KBLCCG significantly outperforms existing adaptive filtering algorithms in terms of prediction accuracy, particularly demonstrating stronger generalization capabilities when dealing with complex chaotic systems. Compared to traditional methods, the kernel broad learning strategy markedly enhances the feature capturing and modeling effectiveness of chaotic time series, further validating the method's efficacy and robustness in nonlinear time series prediction. The KBLCCG algorithm not only exhibits superior predictive capabilities in complex non-Gaussian noise environments but also provides an innovative solution for handling the nonlinear and chaotic characteristics of time series prediction.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130234"},"PeriodicalIF":5.5,"publicationDate":"2025-04-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869454","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}

{"title":"Privacy-preserving average consensus for second-order discrete-time multi-agent systems","authors":"Jie Wang,&nbsp;Na Huang,&nbsp;Yun Chen,&nbsp;Qiang Lu","doi":"10.1016/j.neucom.2025.130239","DOIUrl":"10.1016/j.neucom.2025.130239","url":null,"abstract":"<div><div>This study addresses the privacy-preserving average consensus problem in second-order discrete multi-agent systems under strongly connected and balanced graphs. When both velocity and position states of each agent are measurable, a novel lightweight algorithm is proposed by introducing perturbation signals into the transmitted information. Specifically, the algorithm is divided into two stages. In the initial stage, each agent introduces perturbation signals into its initial position and velocity states during transmission to confound potential attackers. In the subsequent stage, the agents use a standard average consensus algorithm to update their states, ensuring accurate convergence to the average of the initial states. Additionally, further considering the scenario where the velocity state is unavailable for each agent, an improved edge-based perturbation algorithm is introduced. Both algorithms not only effectively prevent the internal honest-but-curious agents from accurately inferring the initial states of other agents, except in the specific case where the curious agent is the sole neighbor of the target agent, but also protect privacy from the external eavesdroppers. Lastly, several numerical examples are conducted to validate the effectiveness of the proposed theoretical approaches.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130239"},"PeriodicalIF":5.5,"publicationDate":"2025-04-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870068","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}

{"title":"Estimating expert prior knowledge from optimization trajectories","authors":"Ville Tanskanen,&nbsp;Petrus Mikkola,&nbsp;Aras Erarslan,&nbsp;Arto Klami","doi":"10.1016/j.neucom.2025.130219","DOIUrl":"10.1016/j.neucom.2025.130219","url":null,"abstract":"<div><div>A recurring task in research is iterative optimization of a process that can be evaluated only by conducting an experiment. Powerful algorithms for assisting this process exist, but they largely ignore the valuable knowledge of expert scientists. We consider a problem within this general scope, not aiming to automate the optimization but instead studying how to infer tacit expert knowledge. This complements the current literature focusing on how such information is used in the optimization process, paying little attention on how the information is obtained. We consider a new formulation where the expertise is inferred by passively observing a human solving an optimization problem, without requiring explicit elicitation techniques. Our solution leverages concepts from Bayesian optimization (BO) commonly used for automating the optimization, but now these tools are used as a theoretical model for the user behavior instead. We assume the expert solves the task approximately in the same manner as a BO algorithm would and solve what kind of prior knowledge about the target function is consistent with the sequence of choices they made. We introduce the problem and a concrete solution, and show that the recovered priors match the true priors in controlled simulated studies. We also empirically evaluate the robustness of the method against violations of the modeling assumptions and demonstrate it on real user data.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"639 ","pages":"Article 130219"},"PeriodicalIF":5.5,"publicationDate":"2025-04-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143863818","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}