Neurocomputing最新文献_第10页

GRIF-PPGNN: Group equality informed Ranking-based Individual Fairness for Privacy-Preserving Graph Neural Network 基于群体平等的基于排名的隐私保护图神经网络个体公平性

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-06 DOI: 10.1016/j.neucom.2025.130583

Xuemin Wang , Yunhui Li , Tianlong Gu , Xuguang Bao , Liang Chang , Guoyong Cai , Tieyuan Liu

{"title":"GRIF-PPGNN: Group equality informed Ranking-based Individual Fairness for Privacy-Preserving Graph Neural Network","authors":"Xuemin Wang , Yunhui Li , Tianlong Gu , Xuguang Bao , Liang Chang , Guoyong Cai , Tieyuan Liu","doi":"10.1016/j.neucom.2025.130583","DOIUrl":"10.1016/j.neucom.2025.130583","url":null,"abstract":"<div><div>Graph Neural Networks (GNNs) play a crucial role in various high-stakes decision-making scenarios due to their superior learning performance and end-to-end design. However, concerns have arisen about GNNs making biased decisions against sensitive subgroups or individuals and their vulnerability to privacy attacks. To address these issues, researchers have proposed methods to enhance both privacy protection and fairness in GNNs. However, existing methods for promoting ranking-based individual fairness only focus on the ranking list with top-K nodes and the optimization tends to promote the fairness of individuals in the majority subgroups. This leads to drastically different levels of individual fairness among groups. Additionally, while some approaches aim for fair and privacy-preserving GNNs, they fail to protect inference privacy, exposing the graph structure during training and testing. We study a novel problem: promoting group equality-informed ranking-based individual fairness (i.e., ensuring both ranking-based individual fairness and fairness across groups) and defending against edge-stealing attacks without exposing the graph structure. To tackle this, we propose GRIF-PPGNN (Group informed Ranking-based Individual Fairness — Privacy-Preserving Graph Neural Network). Our framework comprises three modules: (i) a privacy-preserving module generating perturbed aggregations, (ii) a utility maximization module using these aggregations to train a classifier for downstream tasks, and (iii) a fairness promotion module optimizing neural network parameters to enhance individual fairness and equalize fairness across groups. Comprehensive experiments on real-world datasets demonstrate that GRIF-PPGNN achieves a good balance between group equality-informed individual fairness and model utility in differential private GNNs.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130583"},"PeriodicalIF":5.5,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144242062","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

Text-attributed community detection in complex networks through LLMs and GNNs: A powerful fusion of language and graphs 通过llm和gnn在复杂网络中进行文本属性社区检测：语言和图形的强大融合

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-06 DOI: 10.1016/j.neucom.2025.130573

Sruthi K.S., A. Sreekumar, Kannan Balakrishnan

{"title":"Text-attributed community detection in complex networks through LLMs and GNNs: A powerful fusion of language and graphs","authors":"Sruthi K.S., A. Sreekumar, Kannan Balakrishnan","doi":"10.1016/j.neucom.2025.130573","DOIUrl":"10.1016/j.neucom.2025.130573","url":null,"abstract":"<div><div>This paper introduces a novel framework for community detection in complex networks by considering advanced embedding techniques to integrate textual information associated with nodes and edges with graph structures. We propose a Text-Attributed Graph (TAG) approach, where textual data from nodes and edges, such as book descriptions and user reviews in book recommendation systems, is transformed into semantic embeddings using pre-trained language models (PLMs). Specifically, we employ the latest state-of-the-art embedding models, including E5-Base, variants of BERT (SBERT, DistiBERT, BERT-Base, and BERT-Large), and OpenAI’s <span>text-embedding-3-large</span>, and the cost-effective <span>text-embedding-ada-002</span>, to enrich graph representations with meaningful contextual features as edge embeddings. These embeddings are integrated into graph neural networks (GNNs), enabling the model to exploit structural and textual contexts to improve community detection performance. The integration of textual embeddings and several GNNs in this manner offers a promising performance for enhancing community detection tasks in complex networks, opening new possibilities for applications in recommendation systems, information retrieval, predictive tasks, and beyond.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130573"},"PeriodicalIF":5.5,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144242250","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

Multi-level aggregation based federated few-shot learning 基于多级聚合的联邦少射学习

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-06 DOI: 10.1016/j.neucom.2025.130574

Mingyue Li , Hong Zhu , Erxi Wang , Deyu Chen , Meiyi Xie

{"title":"Multi-level aggregation based federated few-shot learning","authors":"Mingyue Li , Hong Zhu , Erxi Wang , Deyu Chen , Meiyi Xie","doi":"10.1016/j.neucom.2025.130574","DOIUrl":"10.1016/j.neucom.2025.130574","url":null,"abstract":"<div><div>Federated learning (FL) enables clients to train a model collaboratively without data sharing. In real-world applications of FL, data scarcity is a common scenario. Existing research attempts to tackle this issue by combining few-shot learning and FL, which enables clients to learn rich prior knowledge from their local base datasets, thereby accurately predicting new classes with limited training data. However, federated few-shot learning frameworks still face challenges, such as overfitting the training models due to limited local client training data and the difficulty of effectively aggregating prior knowledge caused by non-IID data distribution among clients. We propose a multi-level aggregation based federated few-shot learning framework to address the above issues. By introducing student and teacher levels, the framework achieves fine-grained aggregation. We alleviate the overfitting of client models through a clustering aggregation method at the student level. Moreover, we mitigate the non-IID problem by using a bidirectional communication strategy between the teacher and student levels to achieve knowledge consensus among the models participating in aggregation. Experimental results on benchmark datasets show that our proposed framework outperforms existing methods. Moreover, our approach matches FL-MAML’s time complexity with minimal overall performance degradation, especially in 1-shot settings (average decrease: 0.6%). We also evaluated framework applicability on public datasets, finding robust performance even with unrelated data, consistently outperforming baseline methods. Our experiments demonstrate the effectiveness and efficiency of the proposed framework. It also has high applicability, while various public datasets can provide effective improvements.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130574"},"PeriodicalIF":5.5,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144261951","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

Q&A driven conversational recommendation integrating multiple interest modeling 问答驱动的会话推荐集成了多种兴趣建模

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-06 DOI: 10.1016/j.neucom.2025.130564

Haiping Zhu , Yuqi Sun , Shuwei Che , Yan Chen

{"title":"Q&A driven conversational recommendation integrating multiple interest modeling","authors":"Haiping Zhu , Yuqi Sun , Shuwei Che , Yan Chen","doi":"10.1016/j.neucom.2025.130564","DOIUrl":"10.1016/j.neucom.2025.130564","url":null,"abstract":"<div><div>The conversational recommendation aims to provide users with recommendations in an interaction form of “System Ask-User Respond”. Existing studies rarely consider combining users’ multi-interest at the attribute level for preference modeling, and conduct inefficient conversational recommendation strategy learning, which affects the recommendation performance and conversation performance. To this end, we proposed a Q-A driven conversational recommendation method integrating multiple interest modeling. Specifically, we integrate the user’s positive and negative feedback to model a session dynamic graph, then use the signed graph convolutional network for graph representation learning, and we model multiple interest sequences based on the attention mechanism, then obtain the user interest state representation by fusing of sequence representations to solve the problem of insufficient user preference representation. Besides, we proposed a personalized decision space optimization method to narrow the range of action candidates and train the model with a multi-agent reinforcement learning method integrating hierarchical decision debiasing to solve the problem of poor conversational recommendation strategy learning effect. Experimental results on three public datasets, LastFM, Yelp, and Book, show that compared with existing conversational recommendation methods, our method demonstrates consistent performance improvement across all datasets. In addition, the results of ablation experiments verify the effectiveness of each component in our method.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130564"},"PeriodicalIF":5.5,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144261946","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

Architectures, variants, and performance of neural operators: A comparative review 神经算子的结构、变体和性能：比较回顾

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-06 DOI: 10.1016/j.neucom.2025.130518

Shengjun Liu , Yu Yu , Ting Zhang , Hanchao Liu , Xinru Liu , Deyu Meng

{"title":"Architectures, variants, and performance of neural operators: A comparative review","authors":"Shengjun Liu , Yu Yu , Ting Zhang , Hanchao Liu , Xinru Liu , Deyu Meng","doi":"10.1016/j.neucom.2025.130518","DOIUrl":"10.1016/j.neucom.2025.130518","url":null,"abstract":"<div><div>In recent years, neural operators have emerged as effective alternatives to traditional numerical solvers. They are known for their efficient computation, excellent generalization, and high solving accuracy. Many researchers have shown interest in their design and application. This paper provides a comprehensive summary and analysis of neural operators. We categorize them into three types based on their architecture: deep operator networks (DeepONets), integral kernel operators, and transformer-based neural operators. We then discuss the basic structures and properties of these operator types. Furthermore, we summarize and discuss the various variants and extensions of these three types of neural operators from three directions: (1) operator basis-based neural operator variants; (2) physics-informed neural operator variants; and (3) application of neural operator variants in complex systems. We also analyze the characteristics and performance of different operator methods through numerical experiments. Taking into account these discussions and analyses, we provide perspectives and suggestions regarding the challenges and potential enhancements for different neural operators. This offers valuable guidance and suggestions for the practical application and development of neural operators.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"648 ","pages":"Article 130518"},"PeriodicalIF":5.5,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279897","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

Fixed-time passivity of multi-weighted coupled memristive Cohen–Grossberg neural networks 多重加权耦合记忆性Cohen-Grossberg神经网络的定时无源性

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-05 DOI: 10.1016/j.neucom.2025.130581

Hong-An Tang , Jin-Wei Li , Xiaofang Hu , Shukai Duan , Lidan Wang

引用次数: 0

GPPT: Graph pyramid pooling transformer for visual scene GPPT：图形金字塔池变压器的视觉场景

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-05 DOI: 10.1016/j.neucom.2025.130729

Zhi-Peng Li , Wen-Jian Liu , Xin Sun , Yi-Jie Pan , Valeriya Gribova , Vladimir Fedorovich Filaretov , Anthony G. Cohn , De-Shuang Huang

{"title":"GPPT: Graph pyramid pooling transformer for visual scene","authors":"Zhi-Peng Li , Wen-Jian Liu , Xin Sun , Yi-Jie Pan , Valeriya Gribova , Vladimir Fedorovich Filaretov , Anthony G. Cohn , De-Shuang Huang","doi":"10.1016/j.neucom.2025.130729","DOIUrl":"10.1016/j.neucom.2025.130729","url":null,"abstract":"<div><div>In the field of computer vision, network architectures are critical to the performance of tasks. Vision Graph Neural Network (ViG) has shown remarkable results in handling various vision tasks with their unique characteristics. However, the lack of multi-scale information in ViG limits its expressive capability. To address this challenge, we propose a Graph Pyramid Pooling Transformer (GPPT), which aims to enhance the performance of the model by introducing multi-scale feature learning. The core advantage of GPPT is its ability to effectively capture and fuse feature information at different scales. Specifically, it first generates multi-level pooled graphs using a graph pyramid pooling structure. Next, it encodes features at each scale using a weight-shared Graph Convolutional Neural Network (GCN). Then, it enhances information exchange across scales through a cross-scale feature fusion mechanism. Finally, it captures long-range node dependencies using a transformer module. The experimental results demonstrate that GPPT achieves exceptional performance across various visual scenes, including image classification, and object detection, highlighting its generality and validity.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130729"},"PeriodicalIF":5.5,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144231433","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

Damped sliding based mining of average utility-driven sequential patterns over uncertain data streams 基于阻尼滑动的不确定数据流上平均效用驱动序列模式挖掘

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-05 DOI: 10.1016/j.neucom.2025.130669

Ruihua Zhang, Meng Han, Feifei He, Fanxing Meng, Chunpeng Li

{"title":"Damped sliding based mining of average utility-driven sequential patterns over uncertain data streams","authors":"Ruihua Zhang, Meng Han, Feifei He, Fanxing Meng, Chunpeng Li","doi":"10.1016/j.neucom.2025.130669","DOIUrl":"10.1016/j.neucom.2025.130669","url":null,"abstract":"<div><div>Stream pattern mining has attracted significant attention due to its broad application prospects. However, existing algorithms face two major challenges in practical applications: First, current methods cannot effectively handle uncertain data streams, limiting their applicability. Second, the temporal nature of data streams means that recent data is typically more valuable than historical data, yet existing algorithms fail to adequately account for the temporal sensitivity of data streams, lacking an effective mechanism to distinguish between new and old data, which compromises the practicality of mining results. To address these issues, this paper proposes a High Average-Utility Sequential Pattern mining algorithm for Uncertain Data Streams (HAUSP_UDS). The algorithm employs a probability model to quantify data uncertainty and incorporates an average utility measure to eliminate the impact of sequence length on mining results. Additionally, a sliding window-based time decay mechanism is designed to dynamically adjust data weights, reflecting the temporal value of sequences. To improve mining efficiency, a Dynamic Sequence List (DSL) structure is introduced to efficiently store and compute utility information, thereby accelerating the calculation of utility values and upper bounds. Experiments on both real-world and synthetic datasets demonstrate that, compared to state-of-the-art algorithms, the proposed method achieves significant improvements in recall, precision, time-space efficiency, and scalability.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130669"},"PeriodicalIF":5.5,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144242013","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

Heterogeneous agents trajectory prediction with dynamic interaction relational reasoning 基于动态交互关系推理的异构智能体轨迹预测

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-04 DOI: 10.1016/j.neucom.2025.130543

Nianwen Ning , Shihan Tian , Hengji Li , Wei Li , Chang Liu , Yi Zhou , XiaoZhi Gao

{"title":"Heterogeneous agents trajectory prediction with dynamic interaction relational reasoning","authors":"Nianwen Ning , Shihan Tian , Hengji Li , Wei Li , Chang Liu , Yi Zhou , XiaoZhi Gao","doi":"10.1016/j.neucom.2025.130543","DOIUrl":"10.1016/j.neucom.2025.130543","url":null,"abstract":"<div><div>Accurate trajectory prediction for different types of agents in complex environments is crucial for enabling safe navigation planning. However, current trajectory prediction methods usually ignore the fact that the behaviors of different types of traffic participants exhibit first-order discontinuities. For example, at intersections, the movement behavior can abruptly shift between stopping, going straight, turning right, or turning left, due to the frequent interactions occurring between the agents and the constraints imposed by traffic rules. Their behavior is directly affected by their interactions with the surrounding agents and the environment. To address these challenges, we propose a trajectory prediction method for heterogeneous agents with dynamic interaction relational reasoning. We utilize type-specific encoders to extract dynamic features of agents from their historical states. Interactions between heterogeneous agents are abstracted as heterogeneous graphs with directed edge features, then processed by a novel graph attention network with dynamic relational reasoning designed to extract spatio-temporal interaction features. To capture dynamic interactions, the graph is evolved into a topologically and representationally dynamic graph. Spatial interaction discontinuities are handled by reconstructing subgraphs for different agents with dynamic and changeable features. Furthermore, to reason about interactions, a two-element relational representation is proposed to obtain dynamic relational reasoning. Finally, we conduct a validation test of the proposed model utilizing real-world datasets. The experimental results from different aspects demonstrate that our method can effectively capture the dynamic interaction features between heterogeneous agents, realize the trajectory prediction of heterogeneous agents, and achieve state-of-the-art performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130543"},"PeriodicalIF":5.5,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144261513","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

Infrared small target detection using the global low-rank and local smoothness coupled representation with local structure 红外小目标检测采用全局低秩和局部光滑耦合表示与局部结构

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-06-04 DOI: 10.1016/j.neucom.2025.130546

Junying Li, Xiaorong Hou, Yajian Zeng

{"title":"Infrared small target detection using the global low-rank and local smoothness coupled representation with local structure","authors":"Junying Li, Xiaorong Hou, Yajian Zeng","doi":"10.1016/j.neucom.2025.130546","DOIUrl":"10.1016/j.neucom.2025.130546","url":null,"abstract":"<div><div>Infrared small target detection is crucial in both military and civilian applications. However, existing low-rank and sparse decomposition (LRSD) methods often suffer from noise residues caused by inaccurate background estimation. This is because the low-rank and smoothness of the background exhibit inherent coupling properties. It is usually difficult to accurately fit complicated background using a single regularization term or their additive hybrid model. This paper tackles this issue by proposing a coupled tensor model that incorporates global low-rank and local smoothness. Furthermore, to suppress potential “grid artifacts”, which are usually brought on by the infrared device’s pixel array characteristics and total variation, another regularization term that focuses on the minimum absolute structure of the tensor’s gradient in the local region is constructed. The proposed model is then solved using an optimization framework based on the alternating direction method of multipliers (ADMM). Finally, comparative experiments on three public datasets demonstrate that the proposed model outperforms existing state-of-the-art LRSD methods in terms of suppressing complicated background and sparse “grid artifacts”.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"647 ","pages":"Article 130546"},"PeriodicalIF":5.5,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144221970","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