IEEE transactions on neural networks and learning systems最新文献_第3页

Learning Counterfactual Fair Representation Under Covariate Shift via Reflux 通过回流学习协变量移位下的反事实公平表征

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-06 DOI: 10.1109/tnnls.2025.3613549

Yiliang Xia, Xiaohang Zhang, Zhengren Li

引用次数: 0

Theoretical Convergence Analysis and Initialization Comparisons of Deep Soft-Thresholding Networks 深度软阈值网络的理论收敛分析及初始化比较

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-06 DOI: 10.1109/tnnls.2025.3614196

Chunyan Xiong, Mengxue Zhang, Tong Wei, Yihui Huang, Jian Cao, Qingrui Cai, Zhong Chen, Xiaobo Qu

引用次数: 0

Hard Sample Mining: A New Paradigm of Efficient and Robust Model Training 硬样本挖掘：高效鲁棒模型训练的新范式

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-06 DOI: 10.1109/tnnls.2025.3610948

Lei Liu, Yunji Liang, Xiaokai Yan, Luwen Huangfu, Sagar Samtani, Zhiwen Yu, Yanyong Zhang, Daniel D. Zeng

引用次数: 0

FedMPS: Federated Learning in a Synergy of Multi-Level Prototype-Based Contrastive Learning and Soft Label Generation FedMPS：基于多级原型的对比学习和软标签生成协同的联邦学习

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-06 DOI: 10.1109/tnnls.2025.3611832

Wenxin Yang, Xingchen Hu, Xiubin Zhu, Rouwan Wu, Witold Pedrycz, Xinwang Liu, Jincai Huang

引用次数: 0

Multigranularity Deep Graph Convolutional Neural Network Node Clustering Leveraging Spatial Information 利用空间信息的多粒度深度图卷积神经网络节点聚类

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-06 DOI: 10.1109/tnnls.2025.3615830

Bin Yu, Haibo Yang

引用次数: 0

FedNK-RF: Federated Kernel Learning With Heterogeneous Data and Optimal Rates FedNK-RF：基于异构数据和最优速率的联邦核学习

IF 10.4 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-03 DOI: 10.1109/tnnls.2025.3612728

Xuning Zhang, Jian Li, Rong Yin, Weiping Wang

引用次数: 0

HVLF: A Holistic Visual Localization Framework Across Diverse Scenes. HVLF：跨不同场景的整体视觉定位框架。

IF 8.9 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-01 DOI: 10.1109/TNNLS.2025.3580405

Kun Dai, Zhiqiang Jiang, Fuyuan Qiu, Dedong Liu, Tao Xie, Ke Wang, Ruifeng Li, Lijun Zhao

{"title":"HVLF: A Holistic Visual Localization Framework Across Diverse Scenes.","authors":"Kun Dai, Zhiqiang Jiang, Fuyuan Qiu, Dedong Liu, Tao Xie, Ke Wang, Ruifeng Li, Lijun Zhao","doi":"10.1109/TNNLS.2025.3580405","DOIUrl":"10.1109/TNNLS.2025.3580405","url":null,"abstract":"Recently, integrating the multitask learning (MTL) paradigm into scene coordinate regression (SCoRe) techniques has achieved significant success in visual localization tasks. However, the feature extraction ability of existing frameworks is inherently constrained by the rigid weight activation strategy, which prevents each layer from concurrently capturing scene-universal features across diverse scenes and scene-particular attributes unique to each individual scene. In addition, the straightforward network architecture further exacerbates the issue of insufficient feature representation. To address these limitations, we introduce HVLF, a holistic framework that ensures flexible identification of both scene-universal and scene-particular attributes while integrating various attention mechanisms to enhance feature representation effectively. Technically, for the first issue, HVLF proposes a soft weight activation strategy (SWAS) equipped with polyhedral convolution to concurrently optimize scene-shared and scene-specific weights within each layer, which facilitates sufficient discernment of both scene-universal features and scene-particular attributes, thereby boosting the network's capability for comprehensive scene perception. For the second issue, HVLF introduces a mixed attention perception module (MAPM) that incorporates channelwise, spatialwise, and elementwise attention mechanisms to perform multilevel feature fusion, hence extracting discriminative features to regress precise scene coordinates. Extensive experiments on indoor and outdoor datasets prove that HVLF realizes impressive localization performance. In addition, experiments conducted on 3-D object detection and feature matching tasks prove that the two proposed techniques are universal and can be seamlessly inserted into other methods.","PeriodicalId":13303,"journal":{"name":"IEEE transactions on neural networks and learning systems","volume":"PP ","pages":"18859-18873"},"PeriodicalIF":8.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144560031","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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

Utilizing TOP2 Class for Hybrid Decision-Making to Enhance TOP1 Accuracy of Ensemble Models. 利用TOP2类进行混合决策提高集成模型的TOP1精度。

IF 8.9 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-01 DOI: 10.1109/TNNLS.2025.3579732

Jiqing Li, Zhendong Yin, Dasen Li, Yanlong Zhao

{"title":"Utilizing TOP2 Class for Hybrid Decision-Making to Enhance TOP1 Accuracy of Ensemble Models.","authors":"Jiqing Li, Zhendong Yin, Dasen Li, Yanlong Zhao","doi":"10.1109/TNNLS.2025.3579732","DOIUrl":"10.1109/TNNLS.2025.3579732","url":null,"abstract":"In the domain of deep learning for visual tasks, ensemble models combine several less accurate models to form a more precise composite model, improving overall performance. Traditionally, majority voting and average probabilities have been the main decision-making techniques in ensemble learning, focusing only on the TOP1 Class of base models, hence overlooking other significant information. This article introduces a new algorithm, TOP2 hybrid decision (TOP2 HD), which enhances the TOP1 accuracy of the ensemble model. TOP2 HD categorizes base models into hierarchies based on their TOP1 Class and uses the TOP2 Class for ranking, leading to better performance. Extensive experiments across various models and datasets demonstrate that TOP2 HD not only surpasses traditional ensemble methods, such as majority voting, average probabilities, and stacking, but also exceeds many of the latest ensemble strategies in the image domain. In addition, our experiments revealed a functional relationship between the test accuracy of the ensemble model and the number of base models. This enables us to predict the upper limit of the ensemble model's performance using only a fraction of the models, providing a crucial reference for the performance after the deployment of the ensemble model.","PeriodicalId":13303,"journal":{"name":"IEEE transactions on neural networks and learning systems","volume":"PP ","pages":"18765-18779"},"PeriodicalIF":8.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144560032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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

Asynchronous Boundary Stabilization of Stochastic Markovian Reaction-Diffusion Neural Networks With Mode-Dependent Delays. 具有模相关延迟的随机马尔可夫反应-扩散神经网络的异步边界镇定。

IF 8.9 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-01 DOI: 10.1109/TNNLS.2025.3574214

Xin-Xin Han, Kai-Ning Wu, Xin Yuan

引用次数: 0

Fed-HeLLo: Efficient Federated Foundation Model Fine-Tuning With Heterogeneous LoRA Allocation. Fed-HeLLo：异构LoRA分配的高效联邦基础模型微调。

IF 8.9 1区计算机科学

IEEE transactions on neural networks and learning systems Pub Date : 2025-10-01 DOI: 10.1109/TNNLS.2025.3580495

Zikai Zhang, Ping Liu, Jiahao Xu, Rui Hu

{"title":"Fed-HeLLo: Efficient Federated Foundation Model Fine-Tuning With Heterogeneous LoRA Allocation.","authors":"Zikai Zhang, Ping Liu, Jiahao Xu, Rui Hu","doi":"10.1109/TNNLS.2025.3580495","DOIUrl":"10.1109/TNNLS.2025.3580495","url":null,"abstract":"Federated learning (FL) has recently been used to collaboratively fine-tune foundation models (FMs) across multiple clients. Notably, federated low-rank adaptation (LoRA)-based fine-tuning methods have recently gained attention, which allows clients to fine-tune FMs with a small portion of trainable parameters locally. However, most existing methods do not account for the heterogeneous resources of clients or lack an effective local training strategy to maximize global fine-tuning performance under limited resources. In this work, we propose federated LoRA-based fine-tuning framework with heterogeneous LoRA allocation (Fed-HeLLo), a novel federated LoRA-based fine-tuning framework that enables clients to collaboratively fine-tune an FM with different local trainable LoRA layers. To ensure its effectiveness, we develop several heterogeneous LoRA allocation (HLA) strategies that adaptively allocate local trainable LoRA layers based on clients' resource capabilities and the layer importance. Specifically, based on the dynamic layer importance, we design a Fisher information matrix score-based HLA (FIM-HLA) that leverages dynamic gradient norm information. To better stabilize the training process, we consider the intrinsic importance of LoRA layers and design a geometrically defined HLA (GD-HLA) strategy. It shapes the collective distribution of trainable LoRA layers into specific geometric patterns, such as triangle, inverted triangle, bottleneck, and uniform. Moreover, we extend GD-HLA into a randomized version, named randomized GD-HLA (RGD-HLA), for enhanced model accuracy with randomness. By codesigning the proposed HLA strategies, we incorporate both the dynamic and intrinsic layer importance into the design of our HLA strategy. To thoroughly evaluate our approach, we simulate various complex federated LoRA-based fine-tuning settings using five datasets and three levels of data distributions ranging from independent identically distributed (i.i.d.) to extreme non-i.i.d. The experimental results demonstrate the effectiveness and efficiency of Fed-HeLLo with the proposed HLA strategies. The code is available at https://github.com/ TNI-playground/Fed_HeLLo.","PeriodicalId":13303,"journal":{"name":"IEEE transactions on neural networks and learning systems","volume":"PP ","pages":"17556-17569"},"PeriodicalIF":8.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144583809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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