Information Sciences最新文献_第5页

Resilient tracking control of UAV with event-triggered communication against stochastic DoS attacks and faults 针对随机DoS攻击和故障的事件触发通信无人机弹性跟踪控制

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-03 DOI: 10.1016/j.ins.2026.123074

Minrui Fu, Ziquan Yu

This article investigates a resilient control strategy for an unmanned aerial vehicle (UAV) under an event-triggered communication (ETC) mechanism to simultaneously resist stochastic denial-of-service (DoS) attacks and faults. Firstly, the network health state during stochastic DoS attacks is modeled as a Markov process by using random variables to represent both normal and interrupted periods. Secondly, an observer incorporating a radial basis function neural network (RBFNN) is developed to estimate unknown components caused by faults and uncertain dynamics. Then, auxiliary control laws are proposed for fault compensation. Furthermore, a resilient controller is designed to enhance system resilience, along with a dynamic ETC mechanism that adaptively adjusts the trigger frequency. Moreover, a stochastic hybrid system model of the UAV is constructed to better incorporate the continuous states, jump states and the random inputs. Within this hybrid system framework, it is proved that the tracking error remains Lagrange stable in probability and Lyapunov stable in probability. Finally, the feasibility and efficiency of the proposed scheme are validated through a hardware-in-the-loop experiment using the open-source flight autopilot Pixhawk

^{®}

6C.

本文研究了一种事件触发通信（ETC）机制下的无人机弹性控制策略，以同时抵抗随机拒绝服务（DoS）攻击和故障。首先，将随机DoS攻击时的网络健康状态建模为马尔可夫过程，使用随机变量表示正常和中断时间段；其次，提出了一种基于径向基函数神经网络（RBFNN）的观测器，用于估计由故障和不确定动态引起的未知分量；然后，提出了辅助控制律进行故障补偿。此外，设计了弹性控制器来增强系统的弹性，以及自适应调整触发频率的动态ETC机制。在此基础上，构建了无人机的随机混合系统模型，以更好地融合连续状态、跳跃状态和随机输入。在此混合系统框架下，证明了跟踪误差在概率上保持拉格朗日稳定，在概率上保持李雅普诺夫稳定。最后，利用开源飞行自动驾驶仪Pixhawk®6C进行了硬件在环实验，验证了该方案的可行性和效率。

{"title":"Resilient tracking control of UAV with event-triggered communication against stochastic DoS attacks and faults","authors":"Minrui Fu, Ziquan Yu","doi":"10.1016/j.ins.2026.123074","DOIUrl":"10.1016/j.ins.2026.123074","url":null,"abstract":"<div><div>This article investigates a resilient control strategy for an unmanned aerial vehicle (UAV) under an event-triggered communication (ETC) mechanism to simultaneously resist stochastic denial-of-service (DoS) attacks and faults. Firstly, the network health state during stochastic DoS attacks is modeled as a Markov process by using random variables to represent both normal and interrupted periods. Secondly, an observer incorporating a radial basis function neural network (RBFNN) is developed to estimate unknown components caused by faults and uncertain dynamics. Then, auxiliary control laws are proposed for fault compensation. Furthermore, a resilient controller is designed to enhance system resilience, along with a dynamic ETC mechanism that adaptively adjusts the trigger frequency. Moreover, a stochastic hybrid system model of the UAV is constructed to better incorporate the continuous states, jump states and the random inputs. Within this hybrid system framework, it is proved that the tracking error remains Lagrange stable in probability and Lyapunov stable in probability. Finally, the feasibility and efficiency of the proposed scheme are validated through a hardware-in-the-loop experiment using the open-source flight autopilot Pixhawk<span><math><msup><mspace></mspace><mi>®</mi></msup></math></span> 6C.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123074"},"PeriodicalIF":6.8,"publicationDate":"2026-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940334","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

Deep multi-view clustering based on cross-mutual information 基于互信息的深度多视图聚类

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-03 DOI: 10.1016/j.ins.2026.123070

Yong Wang , Yifan Zhang , Guifu Lu , Cuiyun Gao

Deep multi-view clustering is a data analysis method based on deep learning technology to obtain feature information from multi-source data to serve downstream clustering tasks. Following the success of deep learning in latent representation, scholarly interest in DMVC has proliferated over the past few years. However, existing DMVC methods usually directly fuse feature information obtained from different views to obtain a shared representation matrix for downstream clustering tasks. Existing DMVC methods often ignore the inherent privacy characteristics of each view. Direct feature merging may cause these unique view-specific elements to prevail in the final embedding, thus affecting the accuracy of the clustering. Motivated by the need to surmount these limitations, this paper introduces DMVCIF, a deep multi-view algorithm built upon the principle of feature mutual information. Specifically, we unify the two tasks of reducing intra-feature private information and feature fusion within the mutual information framework, achieving these two goals in the training and optimization process, and finally obtaining high-quality shared feature representation matrices for downstream clustering tasks. Empirical evaluation on seven benchmark datasets shows that the proposed method consistently outperforms the existing comparison baseline algorithms. In addition, we designed experiments based on large-scale data samples in the presence of noise to verify the potential of the model when dealing with large datasets. The source code for our proposed DMVCIF is publicly accessible at github.com/snothingtosay/DMVCIF.

深度多视图聚类是一种基于深度学习技术，从多源数据中获取特征信息，服务于下游聚类任务的数据分析方法。随着深度学习在潜在表征方面的成功，学术界对DMVC的兴趣在过去几年中激增。然而，现有的DMVC方法通常是直接融合不同视图的特征信息，得到一个共享的表示矩阵，用于下游聚类任务。现有的DMVC方法经常忽略每个视图的固有隐私特征。直接特征合并可能导致这些独特的特定于视图的元素在最终嵌入中占上风，从而影响聚类的准确性。基于克服这些局限性的需要，本文提出了基于特征互信息原理的深度多视图算法DMVCIF。具体而言，我们将互信息框架内特征内部私有信息减少和特征融合两项任务统一起来，在训练和优化过程中实现这两个目标，最终获得用于下游聚类任务的高质量共享特征表示矩阵。对7个基准数据集的实证评估表明，本文提出的方法始终优于现有的比较基线算法。此外，我们设计了基于存在噪声的大规模数据样本的实验，以验证模型在处理大型数据集时的潜力。我们建议的DMVCIF的源代码可以在github.com/snothingtosay/DMVCIF上公开访问。

{"title":"Deep multi-view clustering based on cross-mutual information","authors":"Yong Wang , Yifan Zhang , Guifu Lu , Cuiyun Gao","doi":"10.1016/j.ins.2026.123070","DOIUrl":"10.1016/j.ins.2026.123070","url":null,"abstract":"<div><div>Deep multi-view clustering is a data analysis method based on deep learning technology to obtain feature information from multi-source data to serve downstream clustering tasks. Following the success of deep learning in latent representation, scholarly interest in DMVC has proliferated over the past few years. However, existing DMVC methods usually directly fuse feature information obtained from different views to obtain a shared representation matrix for downstream clustering tasks. Existing DMVC methods often ignore the inherent privacy characteristics of each view. Direct feature merging may cause these unique view-specific elements to prevail in the final embedding, thus affecting the accuracy of the clustering. Motivated by the need to surmount these limitations, this paper introduces DMVCIF, a deep multi-view algorithm built upon the principle of feature mutual information. Specifically, we unify the two tasks of reducing intra-feature private information and feature fusion within the mutual information framework, achieving these two goals in the training and optimization process, and finally obtaining high-quality shared feature representation matrices for downstream clustering tasks. Empirical evaluation on seven benchmark datasets shows that the proposed method consistently outperforms the existing comparison baseline algorithms. In addition, we designed experiments based on large-scale data samples in the presence of noise to verify the potential of the model when dealing with large datasets. The source code for our proposed DMVCIF is publicly accessible at <span><span>github.com/snothingtosay/DMVCIF</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123070"},"PeriodicalIF":6.8,"publicationDate":"2026-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940214","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

A two-stage self-supervised learning framework for breast cancer detection with multi-scale vision transformers 基于多尺度视觉变换的乳腺癌检测两阶段自监督学习框架

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-03 DOI: 10.1016/j.ins.2025.123061

Shahriar Mohammadi, Mohammad Ahmadi Livani

Breast cancer detection through mammography remains a cornerstone of early diagnosis, yet the limited availability of large, expertly annotated datasets poses a significant challenge for developing robust AI models. To address this data scarcity, we propose a novel Two-Stage Self-Supervised Learning (TSSL) framework named TSSL-MSViT, which utilizes a Multi-Scale Vision Transformer (MSViT) to learn data-efficient mammographic representations. In Stage 1, the MSViT backbone is pretrained using a dual-objective strategy that integrates Multi-Scale Masked Reconstruction (MS-MR) and Cross-Scale Contrastive Learning (CS-C). Unlike prior single-task SSL pipelines, MS-MR captures fine- and coarse-grained structures, while CS-C explicitly aligns multi-resolution and multi-view (CC/MLO) semantics, yielding representations that are simultaneously hierarchical and view-consistent. This synergistic design provides a principled foundation—beyond empirical gains—for learning stable and transferable mammographic features from unlabeled data. In Stage 2, the pretrained MSViT backbone is fine-tuned with limited labeled data for breast-level classification. Comprehensive experiments on the CBIS-DDSM and INbreast datasets demonstrate that TSSL-MSViT consistently outperforms both Convolutional Neural Network (CNN) and Vision Transformer baselines. The model achieves state-of-the-art AUCs of 0.967 (CBIS-DDSM) and 0.972 (INbreast), significantly surpassing the Swin Transformer and other leading architectures. These results highlight the effectiveness of combining multi-scale feature modeling with self-supervised representation learning for data-efficient, generalizable, and accurate mammographic analysis. The proposed framework establishes a strong foundation for future AI-driven diagnostic systems, reducing dependence on extensive expert annotations while enhancing clinical reliability.

通过乳房x光检查检测乳腺癌仍然是早期诊断的基石，但大型、专业注释数据集的可用性有限，这对开发强大的人工智能模型构成了重大挑战。为了解决这一数据稀缺问题，我们提出了一种新的两阶段自监督学习（TSSL）框架，名为TSSL-MSViT，它利用多尺度视觉转换器（MSViT）来学习数据高效的乳房x线摄影表示。在第一阶段，使用融合了多尺度掩膜重建（MS-MR）和跨尺度对比学习（CS-C）的双目标策略对MSViT主干进行预训练。与之前的单任务SSL管道不同，MS-MR捕获细粒度和粗粒度结构，而CS-C显式地对齐多分辨率和多视图（CC/MLO）语义，产生分层和视图一致的表示。这种协同设计为从未标记的数据中学习稳定和可转移的乳房x线摄影特征提供了一个原则基础-超越经验收益。在第二阶段，预训练的MSViT骨干与有限的标记数据进行微调，用于乳房水平分类。在CBIS-DDSM和INbreast数据集上的综合实验表明，tsll - msvit始终优于卷积神经网络（CNN）和Vision Transformer基线。该模型实现了0.967 （CBIS-DDSM）和0.972 （INbreast）的最先进的auc，大大超过了Swin Transformer和其他领先的架构。这些结果强调了将多尺度特征建模与自监督表示学习相结合，用于数据高效、可推广和准确的乳房x线检查分析的有效性。提出的框架为未来人工智能驱动的诊断系统奠定了坚实的基础，减少了对大量专家注释的依赖，同时提高了临床可靠性。

{"title":"A two-stage self-supervised learning framework for breast cancer detection with multi-scale vision transformers","authors":"Shahriar Mohammadi, Mohammad Ahmadi Livani","doi":"10.1016/j.ins.2025.123061","DOIUrl":"10.1016/j.ins.2025.123061","url":null,"abstract":"<div><div>Breast cancer detection through mammography remains a cornerstone of early diagnosis, yet the limited availability of large, expertly annotated datasets poses a significant challenge for developing robust AI models. To address this data scarcity, we propose a novel Two-Stage Self-Supervised Learning (TSSL) framework named TSSL-MSViT, which utilizes a Multi-Scale Vision Transformer (MSViT) to learn data-efficient mammographic representations. In Stage 1, the MSViT backbone is pretrained using a dual-objective strategy that integrates Multi-Scale Masked Reconstruction (MS-MR) and Cross-Scale Contrastive Learning (CS-C). Unlike prior single-task SSL pipelines, MS-MR captures fine- and coarse-grained structures, while CS-C explicitly aligns multi-resolution and multi-view (CC/MLO) semantics, yielding representations that are simultaneously hierarchical and view-consistent. This synergistic design provides a principled foundation—beyond empirical gains—for learning stable and transferable mammographic features from unlabeled data. In Stage 2, the pretrained MSViT backbone is fine-tuned with limited labeled data for breast-level classification. Comprehensive experiments on the CBIS-DDSM and INbreast datasets demonstrate that TSSL-MSViT consistently outperforms both Convolutional Neural Network (CNN) and Vision Transformer baselines. The model achieves state-of-the-art AUCs of 0.967 (CBIS-DDSM) and 0.972 (INbreast), significantly surpassing the Swin Transformer and other leading architectures. These results highlight the effectiveness of combining multi-scale feature modeling with self-supervised representation learning for data-efficient, generalizable, and accurate mammographic analysis. The proposed framework establishes a strong foundation for future AI-driven diagnostic systems, reducing dependence on extensive expert annotations while enhancing clinical reliability.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123061"},"PeriodicalIF":6.8,"publicationDate":"2026-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940384","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

Exploring prompt engineering with instructive shots to enhance LLMs as a recommender system 探索提示工程与指导镜头，以加强法学硕士推荐系统

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-03 DOI: 10.1016/j.ins.2025.123064

Jianye Xie , Yunpeng Sun , Tiyao Liu

With the development and remarkable achievements of large language models (LLMs), there has been a substantial increase in interest in the method of Prompt Engineering (PE). PE involves designing specific prompts with a limited number of examples (shots) to improve LLMs’ task understanding and performance. Several studies have attempted to leverage PE techniques in recommendation tasks. However, most existing works rely on randomly sampled shots, which can lead to misleading guidance and poor performance. How to select the most instructive shots in PE techniques remains unexplored, and some advanced PE techniques have yet to be explored in the field of recommender systems (RSs). To address the gaps mentioned above, we propose a matching-based method to select instructive shots, utilizing these selected shots and unexplored PE techniques to enhance LLMs as RSs. We propose a comprehensive catalog of prompt patterns combined with various advanced PE techniques to enhance the capabilities of LLMs in recommendation tasks. We utilize an open-source LLM to conduct extensive experiments on three datasets across two recommendation tasks (item rating, reranking) to evaluate the performance of our method. Additionally, we offer valuable insights into the strengths and limitations of different PE techniques across various recommendation tasks.

随着大型语言模型（large language models, llm）的发展和显著成就，人们对提示工程（Prompt Engineering， PE）方法的兴趣大大增加。PE包括用有限数量的例子（镜头）设计特定的提示，以提高法学硕士的任务理解和表现。一些研究试图在推荐任务中利用PE技术。然而，大多数现有的作品依赖于随机采样的镜头，这可能导致误导性的指导和较差的性能。如何在PE技术中选择最具指导意义的镜头仍然是一个有待探索的问题，一些先进的PE技术在推荐系统（RSs）领域还有待探索。为了解决上述差距，我们提出了一种基于匹配的方法来选择有指导意义的镜头，利用这些选择的镜头和未开发的PE技术来增强llm作为RSs。我们提出了一个综合的提示模式目录，结合各种先进的PE技术来增强llm在推荐任务中的能力。我们利用一个开源的法学硕士在三个数据集上进行了广泛的实验，涉及两个推荐任务（项目评级，重新排名），以评估我们的方法的性能。此外，我们还对不同PE技术在各种推荐任务中的优势和局限性提供了有价值的见解。

{"title":"Exploring prompt engineering with instructive shots to enhance LLMs as a recommender system","authors":"Jianye Xie , Yunpeng Sun , Tiyao Liu","doi":"10.1016/j.ins.2025.123064","DOIUrl":"10.1016/j.ins.2025.123064","url":null,"abstract":"<div><div>With the development and remarkable achievements of large language models (LLMs), there has been a substantial increase in interest in the method of Prompt Engineering (PE). PE involves designing specific prompts with a limited number of examples (shots) to improve LLMs’ task understanding and performance. Several studies have attempted to leverage PE techniques in recommendation tasks. However, most existing works rely on randomly sampled shots, which can lead to misleading guidance and poor performance. How to select the most instructive shots in PE techniques remains unexplored, and some advanced PE techniques have yet to be explored in the field of recommender systems (RSs). To address the gaps mentioned above, we propose a matching-based method to select instructive shots, utilizing these selected shots and unexplored PE techniques to enhance LLMs as RSs. We propose a comprehensive catalog of prompt patterns combined with various advanced PE techniques to enhance the capabilities of LLMs in recommendation tasks. We utilize an open-source LLM to conduct extensive experiments on three datasets across two recommendation tasks (item rating, reranking) to evaluate the performance of our method. Additionally, we offer valuable insights into the strengths and limitations of different PE techniques across various recommendation tasks.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123064"},"PeriodicalIF":6.8,"publicationDate":"2026-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940337","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

BDNet: a deep learning-based point cloud denoising network for Brassica rapa BDNet：基于深度学习的油菜点云去噪网络

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2026.123072

Hanzhe Shi , Zisheng Chen , Junmin Chen , Dongfeng Liu , Huijun Yang

High-quality denoising of Brassica rapa point clouds is an important preprocessing step in agricultural applications, including crop phenotypic analysis and organ segmentation. The traditional methods are extremely sensitive to parameter settings and require fine-tuning. Nevertheless, existing deep learning-based methods primarily focus on CAD models with simple structures. For Brassica rapa point clouds, which exhibit complex structural characteristics, balancing noise removal with structural detail preservation remains challenging. In this paper, a Brassica rapa point cloud denoising network based on point-by-point displacement is proposed, which integrates multi-head attention and local offset features. In particular, it includes the following components: (1) a stacked multi-head attention module to alleviate structural loss and deformation during denoising; (2) a local offset feature module to further enhance denoising performance; (3) a dedicated dataset for Brassica rapa point cloud denoising to evaluate the proposed network. The comparison results show that BDNet realizes an average performance improvement of 10.7 % compared with Pointfilter, 15.0 % compared with PCN, and 12.7 % compared with SDN, indicating superior denoising performance. The proposed method not only improves denoising quality but also effectively retains the structural characteristics of the original Brassica rapa.

油菜点云的高质量去噪是作物表型分析和器官分割等农业应用中的重要预处理步骤。传统方法对参数设置非常敏感，需要进行微调。然而，现有的基于深度学习的方法主要集中在结构简单的CAD模型上。对于具有复杂结构特征的油菜点云，如何平衡去噪与结构细节的保留仍然是一个挑战。本文提出了一种基于逐点位移的油菜点云去噪网络，该网络融合了多头注意力和局部偏移特征。具体而言，它包括以下组成部分：(1)用于减轻去噪过程中结构损失和变形的堆叠多头注意模块；(2)局部偏移特征模块，进一步增强去噪性能；(3)建立油菜点云去噪专用数据集，对所提出的网络进行评价。对比结果表明，BDNet比Pointfilter平均性能提高10.7%，比PCN平均性能提高15.0%，比SDN平均性能提高12.7%，降噪性能优越。该方法不仅提高了去噪质量，而且有效地保留了原始油菜的结构特征。

{"title":"BDNet: a deep learning-based point cloud denoising network for Brassica rapa","authors":"Hanzhe Shi , Zisheng Chen , Junmin Chen , Dongfeng Liu , Huijun Yang","doi":"10.1016/j.ins.2026.123072","DOIUrl":"10.1016/j.ins.2026.123072","url":null,"abstract":"<div><div>High-quality denoising of Brassica rapa point clouds is an important preprocessing step in agricultural applications, including crop phenotypic analysis and organ segmentation. The traditional methods are extremely sensitive to parameter settings and require fine-tuning. Nevertheless, existing deep learning-based methods primarily focus on CAD models with simple structures. For Brassica rapa point clouds, which exhibit complex structural characteristics, balancing noise removal with structural detail preservation remains challenging. In this paper, a Brassica rapa point cloud denoising network based on point-by-point displacement is proposed, which integrates multi-head attention and local offset features. In particular, it includes the following components: (1) a stacked multi-head attention module to alleviate structural loss and deformation during denoising; (2) a local offset feature module to further enhance denoising performance; (3) a dedicated dataset for Brassica rapa point cloud denoising to evaluate the proposed network. The comparison results show that BDNet realizes an average performance improvement of 10.7 % compared with Pointfilter, 15.0 % compared with PCN, and 12.7 % compared with SDN, indicating superior denoising performance. The proposed method not only improves denoising quality but also effectively retains the structural characteristics of the original Brassica rapa.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123072"},"PeriodicalIF":6.8,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940333","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

Generalizing three-way concept lattices for conflict analysis 为冲突分析推广三向概念格

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2025.123053

Jie Zhao

Conflict is a common phenomenon across various domains of life, including interpersonal relationships, business decision-making, social organizations, political parties, and international relations. Conflict analysis aids in resolving disputes among agents and improving their decision-making abilities. This study aims to understand the formation mechanism of conflict and effectively capture conflict-related information in complex decision-making contexts. To this end, we introduce generalized supportive, opposed, and neutral operators derived from agents’ behaviors. By incorporating these three stances, we construct generalized three-way conflict concept lattices to model conflict relationships among agents. Specifically, these generalized conflict concept lattices effectively represent agents’ attitudes in complex decision-making environments. The validity of our approach is further demonstrated through a case study, algorithms, theoretical proofs, and comparative analysis. Compared to existing models, the proposed lattices possess several advantages: they capture more comprehensive conflict information, describe agents’ attitudes in a more nuanced manner, and better explain real-world conflict phenomena.

冲突是生活各个领域的普遍现象，包括人际关系、商业决策、社会组织、政党和国际关系。冲突分析有助于解决代理人之间的纠纷，提高代理人的决策能力。本研究旨在了解复杂决策情境下冲突的形成机制，并有效获取冲突相关信息。为此，我们从智能体的行为中引入了广义的支持、反对和中立算子。通过结合这三种立场，我们构建了广义的三方冲突概念格来模拟agent之间的冲突关系。具体而言，这些广义冲突概念格有效地表征了复杂决策环境下主体的态度。通过案例研究、算法、理论证明和比较分析进一步证明了我们方法的有效性。与现有模型相比，提出的格具有几个优势：它们捕获更全面的冲突信息，以更细致的方式描述代理的态度，并更好地解释现实世界的冲突现象。

{"title":"Generalizing three-way concept lattices for conflict analysis","authors":"Jie Zhao","doi":"10.1016/j.ins.2025.123053","DOIUrl":"10.1016/j.ins.2025.123053","url":null,"abstract":"<div><div>Conflict is a common phenomenon across various domains of life, including interpersonal relationships, business decision-making, social organizations, political parties, and international relations. Conflict analysis aids in resolving disputes among agents and improving their decision-making abilities. This study aims to understand the formation mechanism of conflict and effectively capture conflict-related information in complex decision-making contexts. To this end, we introduce generalized supportive, opposed, and neutral operators derived from agents’ behaviors. By incorporating these three stances, we construct generalized three-way conflict concept lattices to model conflict relationships among agents. Specifically, these generalized conflict concept lattices effectively represent agents’ attitudes in complex decision-making environments. The validity of our approach is further demonstrated through a case study, algorithms, theoretical proofs, and comparative analysis. Compared to existing models, the proposed lattices possess several advantages: they capture more comprehensive conflict information, describe agents’ attitudes in a more nuanced manner, and better explain real-world conflict phenomena.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"736 ","pages":"Article 123053"},"PeriodicalIF":6.8,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145981583","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

Exploiting reliable evolving micro-clusters for robust semi-supervised learning on data streams 利用可靠的进化微集群对数据流进行稳健的半监督学习

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2025.123069

Hongliang Wang , Zhonglin Wu , Jinxia Guo , Wei Han , Lei Liu , Qinli Yang , Junming Shao

Traditional semi-supervised learning (SSL) algorithms often heavily depend on assumptions such as clustering and low-density separation. However, these assumptions are frequently violated in complex scenarios, for example, class overlap in feature spaces can cause SSL to perform even worse than using only labeled data. The situation becomes more severe for SSL on evolving data streams as the data distribution changes over time. This makes it more difficult for models to distinguish between closely intertwined classes and effectively adapt to new concepts. In this paper, we propose a novel robust SSL algorithm for evolving data streams with label scarcity by exploiting reliable micro-clusters. To this end, we utilize class disentangled latent representation to learn a reduced, distinguishable and more efficient feature representation of the original streaming data to address class overlap. A set of reliable micro-clusters is dynamically maintained in the representation space to handle and adapt to concept drifts. Finally, reliable instances are selected by considering the reliable properties of micro-clusters and the consistency of an introduced confidence network for online model maintenance. Empirical results on various datasets demonstrate that our method can achieve high performance by effectively exploiting reliable micro-clusters and outperform the existing SSL algorithms.

传统的半监督学习（SSL）算法通常严重依赖于聚类和低密度分离等假设。然而，在复杂的场景中，这些假设经常被违反，例如，特征空间中的类重叠可能导致SSL的性能比仅使用标记数据更差。随着数据分布随时间的变化，SSL在不断发展的数据流上的情况变得更加严重。这使得模型更难以区分紧密交织的类并有效地适应新概念。在本文中，我们提出了一种新的鲁棒SSL算法，该算法利用可靠的微集群来处理标签稀缺性数据流。为此，我们利用类解纠缠的潜在表示来学习原始流数据的简化、可区分和更有效的特征表示，以解决类重叠问题。在表示空间中动态维护一组可靠的微聚类，以处理和适应概念漂移。最后，通过考虑微集群的可靠特性和引入的在线模型维护置信网络的一致性，选择可靠实例。在不同数据集上的实验结果表明，我们的方法可以通过有效地利用可靠的微集群来实现高性能，并且优于现有的SSL算法。

{"title":"Exploiting reliable evolving micro-clusters for robust semi-supervised learning on data streams","authors":"Hongliang Wang , Zhonglin Wu , Jinxia Guo , Wei Han , Lei Liu , Qinli Yang , Junming Shao","doi":"10.1016/j.ins.2025.123069","DOIUrl":"10.1016/j.ins.2025.123069","url":null,"abstract":"<div><div>Traditional semi-supervised learning (SSL) algorithms often heavily depend on assumptions such as clustering and low-density separation. However, these assumptions are frequently violated in complex scenarios, for example, class overlap in feature spaces can cause SSL to perform even worse than using only labeled data. The situation becomes more severe for SSL on evolving data streams as the data distribution changes over time. This makes it more difficult for models to distinguish between closely intertwined classes and effectively adapt to new concepts. In this paper, we propose a novel robust SSL algorithm for evolving data streams with label scarcity by exploiting reliable micro-clusters. To this end, we utilize class disentangled latent representation to learn a reduced, distinguishable and more efficient feature representation of the original streaming data to address class overlap. A set of reliable micro-clusters is dynamically maintained in the representation space to handle and adapt to concept drifts. Finally, reliable instances are selected by considering the reliable properties of micro-clusters and the consistency of an introduced confidence network for online model maintenance. Empirical results on various datasets demonstrate that our method can achieve high performance by effectively exploiting reliable micro-clusters and outperform the existing SSL algorithms.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123069"},"PeriodicalIF":6.8,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940213","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

Actuator fault detection for cyber-physical systems in presence of sparse sensor attacks 存在稀疏传感器攻击的网络物理系统执行器故障检测

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2026.123071

Jianan Zhang , Yining Qian , An-Yang Lu

This paper addresses the actuator fault detection problem under sparse sensor attacks and external disturbances for cyber-physical systems (CPSs). Unlike most studies focusing solely on attacks or faults, this work emphasizes fault detection in attack environments. An attack-resistant fault detection filter is designed to identify actuator faults. To mitigate the impact of attacks on fault detection, an improved switched sliding mode observer (SMO) is proposed, which counteracts attacks by reconstructing the attack signals. Specifically, this observer adaptively selects optimal combined patterns via a dynamic switching mechanism. Furthermore, system asymptotic stability criteria are derived using an augmented Lyapunov function, while

H_{\infty}

performance indices ensure detection robustness under hybrid disturbances. At last, simulation studies demonstrate the effectiveness of the proposed approach in detecting faults subject to sparse sensor attacks.

本文研究了网络物理系统（cps）在稀疏传感器攻击和外部干扰下的执行器故障检测问题。与大多数只关注攻击或故障的研究不同，这项工作强调攻击环境中的故障检测。设计了一种抗攻击故障检测滤波器来识别执行器故障。为了减轻攻击对故障检测的影响，提出了一种改进的切换滑模观测器（SMO），该观测器通过重构攻击信号来抵消攻击。具体来说，该观测器通过动态切换机制自适应选择最优组合模式。利用增广Lyapunov函数导出了系统的渐近稳定性判据，H∞性能指标保证了系统在混合扰动下的检测鲁棒性。最后，仿真研究证明了该方法在检测稀疏传感器攻击下的故障时的有效性。

引用次数: 0

Nonuniform low-light image enhancement based on game-retinex variational and adaptive vector-valued gamma correction 基于game-retinex变分和自适应矢量值伽玛校正的非均匀弱光图像增强

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2025.123063

Wenyang Wei, Xiangchu Feng, Wenhang Song, Tiantian Liu, Weiwei Wang

Non-uniform low-light image enhancement remains challenging due to the coupled estimation of illumination and reflectance in Retinex-based methods, which often leads to over-smoothed illumination or detail loss in reflectance. Moreover, conventional gamma correction with a fixed parameter for the illumination layer fails to balance enhancement across regions with drastically varying brightness. To address these issues, we propose a two-stage framework. First, we formulate Retinex decomposition as a two-player game, in which the illumination and reflectance layers optimize their own utility functions conditioned on each other’s strategy. This game-theoretic interaction, regularized by adaptive fractional-order total variation, enforces mutual guidance between layers through the explicit modeling of their intrinsic coupling, thereby reducing undesirable coupling in the estimation process. Second, we introduce an adaptive vector-valued gamma correction applied to both illumination and reflectance layers, which enhances dark regions without overexposing bright areas while simultaneously compensating for numerical dissipation in texture-rich regions. Experimental results demonstrate both the effectiveness and robustness of the proposed method across diverse lighting conditions.

非均匀弱光图像增强仍然具有挑战性，因为基于视黄素的方法对照度和反射率进行耦合估计，往往导致照度过度平滑或反射率细节丢失。此外，传统的具有固定参数的照明层伽玛校正无法平衡亮度剧烈变化的区域之间的增强。为了解决这些问题，我们提出了一个分两阶段的框架。首先，我们将Retinex分解描述为一个两人博弈，其中光照层和反射率层根据彼此的策略优化各自的效用函数。这种博弈论的相互作用，通过自适应分数阶总变差进行正则化，通过对其内在耦合的显式建模，加强了层之间的相互引导，从而减少了估计过程中不希望出现的耦合。其次，我们引入了一种适用于照明和反射层的自适应矢量值伽玛校正，它可以增强暗区域而不会过度暴露明亮区域，同时补偿纹理丰富区域的数值耗散。实验结果证明了该方法在不同光照条件下的有效性和鲁棒性。

{"title":"Nonuniform low-light image enhancement based on game-retinex variational and adaptive vector-valued gamma correction","authors":"Wenyang Wei, Xiangchu Feng, Wenhang Song, Tiantian Liu, Weiwei Wang","doi":"10.1016/j.ins.2025.123063","DOIUrl":"10.1016/j.ins.2025.123063","url":null,"abstract":"<div><div>Non-uniform low-light image enhancement remains challenging due to the coupled estimation of illumination and reflectance in Retinex-based methods, which often leads to over-smoothed illumination or detail loss in reflectance. Moreover, conventional gamma correction with a fixed parameter for the illumination layer fails to balance enhancement across regions with drastically varying brightness. To address these issues, we propose a two-stage framework. First, we formulate Retinex decomposition as a two-player game, in which the illumination and reflectance layers optimize their own utility functions conditioned on each other’s strategy. This game-theoretic interaction, regularized by adaptive fractional-order total variation, enforces mutual guidance between layers through the explicit modeling of their intrinsic coupling, thereby reducing undesirable coupling in the estimation process. Second, we introduce an adaptive vector-valued gamma correction applied to both illumination and reflectance layers, which enhances dark regions without overexposing bright areas while simultaneously compensating for numerical dissipation in texture-rich regions. Experimental results demonstrate both the effectiveness and robustness of the proposed method across diverse lighting conditions.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123063"},"PeriodicalIF":6.8,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940339","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

Corrigendum to “CDiffu-DL: Few-shot line drawing extraction from Dunhuang murals via self-supervised pre-training of condition fusion with diffusion model”. [Inf. Sci. 731 (2026) 122931] “CDiffu-DL：基于条件融合扩散模型自监督预训练的敦煌壁画少镜头线条提取”的勘误表。[参考文献731 (2026)122931]

IF 6.8 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2026-01-02 DOI: 10.1016/j.ins.2025.123067

Yuan Ding, Kaijun Wu

引用次数: 0