Knowledge-Based Systems最新文献

{"title":"Scalable tri-factorization guided multi-view subspace clustering","authors":"Guang-Yu Zhang ,&nbsp;Chang-Bin Guan ,&nbsp;Dong Huang ,&nbsp;Zihao Wen ,&nbsp;Chang-Dong Wang ,&nbsp;Lei Xiao","doi":"10.1016/j.knosys.2025.113119","DOIUrl":"10.1016/j.knosys.2025.113119","url":null,"abstract":"<div><div>Anchor-based Multi-view Subspace Clustering (AMSC) has exhibited its outstanding capability in large-scale multi-view clustering. Despite significant progress, previous AMSC approaches still suffer from two limitations. First, they mostly neglect the high-order correlation, which undermines their ability in discovering complex cluster structures. Second, they frequently overlook the potential connection between multi-view dimension reduction and anchor subspace clustering, which affects their robustness to low-quality views. In view of these issues, we present a Scalable Tri-factorization Guided Multi-view Subspace Clustering (ST-MSC) approach. Specifically, the proposed approach seeks to recover the latent sample-anchor relationships in multiple embedded spaces, where the multi-view anchor representations are stacked into a low-rank tensor to enhance their high-order correlations with tri-factorization guidance. Theoretical analysis indicates that the tri-factorization paradigm has inherent relevance with two mutually beneficial tasks, namely, the multi-view dimensionality reduction and the anchor-based multi-view subspace clustering. Furthermore, a simple yet fast algorithm is devised to minimize the objective model, where the latent embedding spaces and the anchor subspace structure can be iteratively updated in a unified manner. Experiments have been conducted to verify the effectiveness and efficiency of our ST-MSC approach in comparison with the advanced approaches.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"312 ","pages":"Article 113119"},"PeriodicalIF":7.2,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143420699","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}

{"title":"AutoPolCNN: A neural architecture search method of convolutional neural network for PolSAR image classification","authors":"Guangyuan Liu ,&nbsp;Yangyang Li ,&nbsp;Yanqiao Chen ,&nbsp;Ronghua Shang ,&nbsp;Licheng Jiao","doi":"10.1016/j.knosys.2025.113122","DOIUrl":"10.1016/j.knosys.2025.113122","url":null,"abstract":"<div><div>Convolutional neural networks (CNNs), as a kind of typical classification model known for good performance, have been utilized to cope with polarimetric synthetic aperture radar (PolSAR) image classification. Nevertheless, the performances of CNNs highly rely on well-designed network architectures and there is no theoretical guarantee on how to design them. As a result, the architectures of CNNs can be only designed by human experts or by trial and error, which makes the architecture design is annoying and time-consuming. So, a neural architecture search (NAS) method of CNN called AutoPolCNN, which can determine the architecture automatically, is proposed in this paper. Specifically, we firstly design the search space which covers the main components of CNNs like convolution and pooling operators. Secondly, considering the fact that the number of layers can also influence the performance of CNN, we propose a super normal module (SNM), which can dynamically adjust the number of network layers according to different datasets in the search stage. Finally, we develop the loss function and the search method for the designed search space. Via AutoPolCNN, preparing the data and waiting for the classification results are enough. Experiments carried out on three PolSAR datasets prove that the architecture can be automatically determined by AutoPolCNN within an hour (<em>at least 10 times faster than existing NAS methods</em>) and has higher overall accuracy (OA) than state-of-the-art (SOTA) PolSAR image classification CNN models.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"312 ","pages":"Article 113122"},"PeriodicalIF":7.2,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403505","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}

{"title":"Escape velocity-based adaptive outlier detection algorithm","authors":"Juntao Yang ,&nbsp;Lijun Yang ,&nbsp;Dongming Tang ,&nbsp;Tao Liu","doi":"10.1016/j.knosys.2025.113116","DOIUrl":"10.1016/j.knosys.2025.113116","url":null,"abstract":"<div><div>Outlier detection is a pivotal technique within the realm of data mining, serving to pinpoint aberrant values nestled within datasets. It has been widely employed across diverse domains, including detection of credit card frauds, identification of seismic activities, and identification of anomalies within image datasets. However, existing approaches still face three shortcomings: (1) they often struggle with the intricacies of parameter selection and the vexing top-n dilemma, (2) they lack in their capacity to discern local outliers, and (3) their algorithmic efficacies markedly wane as datasets burgeon in sample point size and outlier prevalence. In addressing these formidable hurdles, we propose a novel, <strong>E</strong>scape <strong>V</strong>elocity-based adaptive <strong>O</strong>utlier <strong>D</strong>etection algorithm, noted as EVOD. The EVOD algorithm calculates the escape velocity of each data sample point and automatically detects the number of outliers by monitoring peak fluctuations in the growth rate of escape velocities of sample points, thereby solving the top-n problem suffered by existing outlier detection algorithms. Experimental results demonstrate that our algorithm, without requiring manual adjustment of parameters, can simultaneously detect global outliers, local outliers, and outlier clusters. In addition, it maintains a good performance even as the number of sample points and outliers in the dataset increases, particularly for complex manifold datasets.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113116"},"PeriodicalIF":7.2,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377686","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}

{"title":"A cluster chaotic optimization for solving power loss and voltage profiles problems on electrical distribution networks","authors":"Primitivo Diaz,&nbsp;Eduardo H. Haro,&nbsp;Omar Avalos,&nbsp;Nayeli Perez","doi":"10.1016/j.knosys.2025.113145","DOIUrl":"10.1016/j.knosys.2025.113145","url":null,"abstract":"<div><div>The growing demand for electricity poses significant challenges in maintaining a reliable and efficient power supply. Optimal Capacitor Placement (OCP) in electrical engineering addresses this issue by strategically positioning capacitor banks within constrained Radial Distribution Networks (RDNs). Traditional optimization methods often struggle with this problem; alternative approaches, such as metaheuristic algorithms, present promising solutions. Despite advances in optimization techniques, challenges in achieving optimal solutions continue. To address these challenges, recent hybrid computational methods, such as the cluster chaotic optimization (CCO) algorithm, have emerged to enhance stability and robustness in finding optimal solutions. The effectiveness of the CCO algorithm lies in its combination of Evolutionary Computation (EC) and Machine Learning (ML) approaches. These approaches improve the search strategy by leveraging information extracted from the solution landscape, resulting in high performance in discovering optimal solutions. In this context, this work aims to utilize the strengths of the CCO algorithm to solve real-world challenges and evaluate its potential in addressing the OCP. The CCO algorithm was tested on three benchmark RDNs to assess its efficacy. Results were compared with those obtained from classical and recently developed methods and analyzed using non-parametric tests. The findings indicate that the CCO algorithm is competitive and robust in solving the OCP, outperforming similar strategies, and demonstrates its effectiveness in optimizing complex real-world problems in electrical engineering.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113145"},"PeriodicalIF":7.2,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395162","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}

{"title":"TF-Attack: Transferable and fast adversarial attacks on large language models","authors":"Zelin Li ,&nbsp;Kehai Chen ,&nbsp;Lemao Liu ,&nbsp;Xuefeng Bai ,&nbsp;Mingming Yang ,&nbsp;Yang Xiang ,&nbsp;Min Zhang","doi":"10.1016/j.knosys.2025.113117","DOIUrl":"10.1016/j.knosys.2025.113117","url":null,"abstract":"<div><div>With the great advancements in large language models (LLMs), <em>adversarial attacks</em> against LLMs have recently attracted increasing attention. We found that pre-existing adversarial attack methodologies exhibit limited transferability and are notably inefficient, particularly when applied to LLMs. In this paper, we analyze the core mechanisms of previous predominant adversarial attack methods, revealing that (1) the distributions of importance score differ markedly among victim models, restricting the transferability; (2) the sequential attack processes induces substantial time overheads. Based on the above two insights, we introduce a new scheme, named <span>TF-Attack</span>, for <strong>T</strong>ransferable and <strong>F</strong>ast adversarial attacks on LLMs. <span>TF-Attack</span> employs an external LLM as a third-party overseer rather than the victim model to identify critical units within sentences. Moreover, <span>TF-Attack</span> introduces the concept of <em>Importance Level</em>, which allows for parallel substitutions of attacks. We conduct extensive experiments on 6 widely adopted benchmarks, evaluating the proposed method through both automatic and human metrics. Results show that our method consistently surpasses previous methods in transferability and delivers significant speed improvements, up to 10<span><math><mo>×</mo></math></span> faster than earlier attack strategies.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"312 ","pages":"Article 113117"},"PeriodicalIF":7.2,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429898","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}

{"title":"Multi-task dual-level adversarial transfer learning boosted RUL estimation of CNC milling tools","authors":"Pei Wang ,&nbsp;Jinrui Liu ,&nbsp;Jingshuai Qi ,&nbsp;Kesong Zhou ,&nbsp;Hongbo Zhai","doi":"10.1016/j.knosys.2025.113152","DOIUrl":"10.1016/j.knosys.2025.113152","url":null,"abstract":"<div><div>Effectively estimating the remaining useful life (RUL) of milling tools is crucial for intelligent preventive maintenance of CNC milling systems. In this paper, a novel generalized RUL estimation model based on multi-task dual-level adversarial transfer learning with multi-level attention (MTDTL-MA) is proposed for tool RUL prediction with variable working conditions. A multi-task learning structure with multi-level attention is used to predict the wear of each tool face in parallel and capture the max wear of entire tools as a health index for more accurate RUL estimation. Multi-channel encoder-decoder self-attention, multi-gate attention and global-local adversarial transferable attention are integrated to emphasize useful wear-related features, tool face-specific features and transferable features between source and target domains, respectively. A new auxiliary subdomain adversarial domain adaptation and global-local adversarial transferable attention is proposed to form a dual-level adversarial domain adaptation to synergistically improve transfer learning. Both the PHM2010 and Ideahouse dataset (2021) are employed to verify the effectiveness of MTDTL-MA, and the results indicate that the proposed method provides higher RUL prediction accuracy compared to several state-of-the-art methods.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"312 ","pages":"Article 113152"},"PeriodicalIF":7.2,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403507","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}

{"title":"ANOGAT-Sparse-TL: A hybrid framework combining sparsification and graph attention for anomaly detection in attributed networks using the optimized loss function incorporating the Twersky loss for improved robustness","authors":"Wasim Khan ,&nbsp;Nadhem Ebrahim","doi":"10.1016/j.knosys.2025.113144","DOIUrl":"10.1016/j.knosys.2025.113144","url":null,"abstract":"<div><div>In recent years, the identification of abnormalities in attributed networks has become essential for applications including social media analysis, cybersecurity, and financial fraud detection. Unsupervised graph anomaly detection techniques seek to recognize infrequent and anomalous patterns in graph-structured data without the necessity of labelled instances. Conventional methods employing Graph Neural Networks (GNNs) frequently encounter difficulties, especially due to the transmission of noisy edges and the intrinsic intricacy of node interrelations. To overcome these restrictions, we introduce ANOGAT-Sparse-TL, an innovative hybrid framework that integrates graph sparsification and Graph Attention Networks (GAT) with autoencoder-based reconstruction for anomaly detection in attributed networks. The sparsification procedure removes extraneous edges and highlights significant node connections, thereby enhancing computational efficiency and improving anomaly detection efficacy. By including GAT, our model carefully allocates significance to pertinent neighboring nodes, yielding enhanced node embeddings. The autoencoder subsequently reconstructs these embeddings to detect abnormalities via reconstruction errors. Incorporating Tversky Loss in the reconstruction process further improves the robustness of the model by effectively addressing the imbalance between normal and anomalous data, prioritizing the detection of rare anomalies. This optimized loss function allows ANOGAT-Sparse-TL to focus on hard-to-reconstruct instances, which are typically indicative of anomalies, and reduces the impact of noisy data on the model's performance. ANOGAT-Sparse-TL effectively integrates attribute-based and structural anomalies, yielding comprehensive anomaly ratings. Comprehensive studies on the four real-world datasets indicate that our strategy surpasses current state-of-the-art methodologies, with enhanced performance. Moreover, the scalability of our methodology guarantees its relevance to extensive real-world networks, rendering it an adaptable option for diverse graph anomaly detection activities. ANOGAT-Sparse-TL, despite its complexity, maintains computational efficiency and provides substantial improvements in anomaly detection inside attributed networks. Future research may concentrate on enhancing interpretability and broadening generalizability to various network architectures.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113144"},"PeriodicalIF":7.2,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clustering matrix regularization guided hierarchical graph pooling","authors":"Zidong Wang ,&nbsp;Liu Yang ,&nbsp;Tingxuan Chen ,&nbsp;Jun Long","doi":"10.1016/j.knosys.2025.113108","DOIUrl":"10.1016/j.knosys.2025.113108","url":null,"abstract":"<div><div>Hierarchical graph pooling effectively captures hierarchical structural information by iteratively simplifying the input graph into smaller graphs using a pooling function, which has demonstrated superior performance in graph-level tasks. However, existing methods often lack a detailed analysis of the pooling function, leading to issues such as noise, loss of essential information, and difficulties in balancing the retention and removal of graph details. In this paper, we address these challenges from an information theory perspective by analyzing information transmission through the clustering matrix within the pooling function. We introduce a novel approach, CMRGP, which is guided by clustering matrix regularization. This method enhances graph representations by selectively filtering task-relevant information from the input graph to create a compressed yet predictive clustering matrix. Specifically, we incorporate high-frequency information via the graph Laplacian matrix and introduce a dynamic gating mechanism to combine both high- and low-frequency information from graph nodes, improving the predictability of the clustering matrix. Additionally, we employ a noise injection technique, adding multivariate independent Gaussian noise to the clustering matrix to compress information and accurately define node category affiliations. Theoretical validation confirms the effectiveness of our approach. We conduct extensive experiments on datasets spanning social networks, biological proteins, and molecular chemistry, totaling 17,372 sample graphs. CMRGP achieves superior performance in graph-level classification, with an average accuracy improvement of 4.36–8.16% across six public datasets, including increases of 4.36% on DD and 8.16% on NCI1.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113108"},"PeriodicalIF":7.2,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143340037","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}

{"title":"Towards consistency of rule-based explainer and black box model — Fusion of rule induction and XAI-based feature importance","authors":"Michał Kozielski ,&nbsp;Marek Sikora ,&nbsp;Łukasz Wawrowski","doi":"10.1016/j.knosys.2025.113092","DOIUrl":"10.1016/j.knosys.2025.113092","url":null,"abstract":"<div><div>Rule-based models offer a human-understandable representation, i.e. they are interpretable. For this reason, they are used to explain the decisions of non-interpretable complex models, referred to as black box models. The generation of such explanations involves the approximation of a black box model by a rule-based model. To date, however, it has not been investigated whether the rule-based model makes decisions in the same way as the black box model it approximates. Decision making in the same way is understood in this work as the consistency of decisions and the consistency of the most important attributes used for decision making. This study proposes a novel approach ensuring that the rule-based surrogate model mimics the performance of the black box model. The proposed solution performs an explanation fusion involving rule generation and taking into account the feature importance determined by the selected XAI methods for the black box model being explained. The result of the method can be both global and local rule-based explanations. The quality of the proposed solution was verified by extensive analysis on 30 tabular benchmark datasets representing classification problems. Evaluation included comparison with the reference method and an illustrative case study. In addition, the paper discusses the possible pathways for the application of the rule-based approach in XAI and how rule-based explanations, including the proposed method, meet the user perspective and requirements for both content and presentation. The software created and a detailed report containing the full experimental results are available on the GitHub repository (<span><span>https://github.com/ruleminer/FI-rules4XAI</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113092"},"PeriodicalIF":7.2,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395303","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}

{"title":"IFM: Integrating and fine-tuning adversarial examples of recommendation system under multiple models to enhance their transferability","authors":"Fulan Qian ,&nbsp;Yan Cui ,&nbsp;Mengyao Xu ,&nbsp;Hai Chen ,&nbsp;Wenbin Chen ,&nbsp;Qian Xu ,&nbsp;Caihong Wu ,&nbsp;Yuanting Yan ,&nbsp;Shu Zhao","doi":"10.1016/j.knosys.2025.113111","DOIUrl":"10.1016/j.knosys.2025.113111","url":null,"abstract":"<div><div>In black-box attack scenarios on recommendation systems, attackers typically rely on surrogate models to approximate the target model and use them to generate adversarial examples due to a lack of knowledge about the internal mechanisms of the target recommendation model. However, reliance on a single surrogate model often leads to adversarial examples that are prone to overfitting, making them vulnerable to local extremes and limiting their transferability across different models. Moreover, methods generating adversarial examples in flat minimum regions fail to consistently perform across diverse models. To address these limitations, this paper proposes a feature integration and fine-tuning framework, IFM, which aims to reduce the overfitting of adversarial examples and enhance their transferability. IFM captures a wider range of attack features by integrating the knowledge of multiple recommendation models and performs fine-tuning to further improve the transferability of the adversarial examples. Experimental results affirm that our approach markedly enhances the transferability of adversarial examples in recommendation systems over existing state-of-the-art techniques, enabling efficient attacks on recommendation models.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"311 ","pages":"Article 113111"},"PeriodicalIF":7.2,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143379008","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}