Knowledge-Based Systems最新文献

{"title":"Dynamic domain information modulation algorithm for multi-domain sentiment analysis","authors":"Chunyi Yue ,&nbsp;Ang Li","doi":"10.1016/j.knosys.2025.114465","DOIUrl":"10.1016/j.knosys.2025.114465","url":null,"abstract":"<div><div>Multidomain sentiment classification aims to improve model performance constrained by limited labeled data in a single domain by utilizing labeled data from multiple domains. Models that simultaneously train domain classifiers and sentiment classifiers have shown benefits. In this framework, domain classification serves as an auxiliary task, supplying crucial information for sentiment analysis. It is generally assumed that the importance of sentiment classification tasks remains consistent across all domains. By contrast, domain classification tasks exhibit variability because the impact of domain information on sentiment analysis differs among fields. This variability can be managed through adjustable weights or hyperparameters. However, as the number of domains grows, existing hyperparameter optimization algorithms face several challenges, including (1) high computational requirements, (2) convergence difficulties, and (3) increased algorithmic complexity. To efficiently generate the domain-specific information required for sentiment classification, we propose a dynamic information modulation algorithm. Specifically, the training process is divided into two phases. In the first phase, a global modulation factor that controls the proportion of domain classification tasks across all domains is established. In the second phase, we introduce an innovative cross-domain balancing modulation algorithm to refine the domain information embedded in the input text. This refinement is achieved using a gradient- and loss-based method. Experimental results show that our approach consistently enhances performance across most domains, achieving improvements of 0.3–1.0 % on 10 of 16 Amazon domains and 0.5–1.5 % on 3 of 5 Yelp domains, while maintaining performance comparable to baseline models in other domains.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114465"},"PeriodicalIF":7.6,"publicationDate":"2025-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159234","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":"Margin-guided parameter decoupling-consensus framework for federated domain generalization in machinery fault diagnosis","authors":"Linhan Gou,&nbsp;Qikang Li,&nbsp;Baoping Tang,&nbsp;Xiaolong Zhang,&nbsp;Zihao Li,&nbsp;Yonggang Liu","doi":"10.1016/j.knosys.2025.114446","DOIUrl":"10.1016/j.knosys.2025.114446","url":null,"abstract":"<div><div>Federated domain generalization (FDG) as a solution to address the cross-client data heterogeneity problem in privacy-sensitive scenarios has drawn extensive attention in the field of intelligent fault diagnosis of industrial equipment in recent years. Nevertheless, most of the existing FDG-based diagnosis methods rely on client feature distribution alignment or data augmentation strategies, risking data leakage caused by the transmission of deep features and statistical information. To overcome the above-mentioned issues, a margin-guided parameter decoupling-consensus (MGPDC) framework is proposed to decouple the dependence of conventional federated domain generalization methods on features and data distributions and realize the extraction of common knowledge across clients. This framework initially employs a federated meta-learning-driven universal feature extractor to create a transferable shared feature space amidst heterogeneous client data, effectively enhancing the generalization ability of the model for unknown working conditions. Next, a parameter decoupling-consensus synergy (PDCS) mechanism is proposed. In this mechanism, an isolation module is established based on the consistency of parameter updates for parameter decoupling, effectively suppressing model update conflict. Subsequently, an implicit alignment mapping approach is devised for the screened parameters with strong consistency to achieve the extraction of cross-domain common knowledge. Then, an adaptive global margin guidance (AGMG) strategy is proposed to mitigate the interference of the blurred class boundaries during the federated process on common knowledge extraction. Finally, extensive experiments using real wind turbine gearbox data demonstrate the effectiveness and advancement of the MGPDC framework.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114446"},"PeriodicalIF":7.6,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159231","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":"Hypergraph-based CLIP hashing for unsupervised cross-modal retrieval","authors":"Qian Zhang ,&nbsp;Jia-Rui Zhao ,&nbsp;Xiao-Qian Liu ,&nbsp;Yu-Wei Zhan ,&nbsp;Zhen-Duo Chen ,&nbsp;Xin Luo ,&nbsp;Xin-Shun Xu","doi":"10.1016/j.knosys.2025.114508","DOIUrl":"10.1016/j.knosys.2025.114508","url":null,"abstract":"<div><div>With the surge of multi-modal data, how to effectively and efficiently find similar information has become an urgent and important need. Among the existing solutions, unsupervised cross-modal hashing can learn from unlabeled data and provide fast and satisfactory retrieval performance, making it a viable solution. However, existing unsupervised cross-modal hashing methods often inadequately model intricate cross-modal semantic relationships. To bridge this gap, this paper proposes a novel Hypergraph-based CLIP Hashing (HCH). Specifically, HCH utilizes the large-scale visual-language pre-trained model CLIP to extract visual and textual features, and employs a cross-modal Transformer to further enhance semantic fusion among these features. Then, to fully capture the semantic relevance among multi-modal data, we construct a semantic-enhanced similarity matrix and design a mean-based weighting scheme to adjust this matrix. Additionally, we compose a hypergraph convolutional network to further explore high-order semantic information within the input data, leading to more compact and high-quality hash codes. To substantiate HCH’s efficacy, we conducted experiments on three commonly used datasets, confirming its superiority over leading baselines.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114508"},"PeriodicalIF":7.6,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222344","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":"ReCAP2: Rectified and context-aware polarization prompting for robust depth enhancement","authors":"Zhenyu Liu ,&nbsp;Jiatong Xu ,&nbsp;Daxin Liu ,&nbsp;Qide Wang ,&nbsp;Jin Cheng ,&nbsp;Jianrong Tan","doi":"10.1016/j.knosys.2025.114498","DOIUrl":"10.1016/j.knosys.2025.114498","url":null,"abstract":"<div><div>Accurate depth perception is fundamental for numerous computer vision applications, yet depth maps acquired from commodity sensors often suffer from artifacts and inaccuracies, necessitating effective enhancement techniques. Polarization imaging, capturing rich geometric cues robust to illumination variations, offers a promising modality to guide this process. However, effectively integrating these cues within learning-based depth enhancement frameworks remains challenging. Existing methods often overlook the inherent representational gap between depth and polarization features and employ context-agnostic fusion mechanisms, incapable of generating prompts adaptive to cross-modal relationships and local context. To address these limitations, we propose a novel Rectified and Context-Aware Polarization Prompting (ReCAP<span><math><msup><mrow></mrow><mn>2</mn></msup></math></span>) framework for depth enhancement models. The ReCAP<span><math><msup><mrow></mrow><mn>2</mn></msup></math></span> first performs initial feature rectification across both channel and spatial dimensions to bridge the modality gap. Subsequently, it generates fine-grained polarization prompts by leveraging dual-level context: utilizing cross-modal context ensures the prompts encode pertinent inter-modality relationships, while processing spatial neighborhood context yields prompts spatially tailored to regional content. Consequently, these dual-context aware prompts provide precise, adaptive guidance for the foundation model, facilitating more robust depth enhancement. Extensive experiments demonstrate the effectiveness of our method. On the multi-modal HAMMER dataset, our method shows superior accuracy and robustness across diverse sensor types in indoor scenes under both full fine-tuning and prompt tuning settings. Furthermore, cross-domain evaluations on the challenging CroMo dataset validate its strong generalization to outdoor environments.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114498"},"PeriodicalIF":7.6,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222353","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 domain generalized UAV tracking framework via frequency-aware learning and target-aligned data augmentation in complex environments","authors":"Erfeng Liu ,&nbsp;Xinde Li ,&nbsp;Heqing Li ,&nbsp;Guoliang Wu ,&nbsp;Tao Shen","doi":"10.1016/j.knosys.2025.114501","DOIUrl":"10.1016/j.knosys.2025.114501","url":null,"abstract":"<div><div>Unmanned Aerial Vehicle (UAV) tracking plays a critical role in airborne autonomous systems, supporting applications such as disaster response, agricultural monitoring, and military surveillance. However, existing tracking methods often exhibit poor generalization in real-world deployments due to domain shifts between the training and target environments. We propose DGTrack, a novel single-source domain generalization framework for UAV visual tracking. DGTrack integrates a Frequency-Aware Learning (FAL) module that separates and adaptively modulates low- and high-frequency components to reduce stylistic interference while enhancing content representation. In addition, a Target-Aligned Augmentation (TAA) module is introduced to improve source domain diversity through multi-level transformations and to align predictions between original and augmented frames by maximizing mutual information. Extensive experiments on the UAVDT and VisDrone2019 datasets demonstrate that DGTrack achieves superior generalization to unseen domains and consistently outperforms state-of-the-art UAV trackers in single-source settings.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114501"},"PeriodicalIF":7.6,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159228","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":"Energy-Efficient adaptive perception for autonomous driving via lightweight policy learning and simulation-based optimization","authors":"Yanzhan Chen ,&nbsp;Fan Yu ,&nbsp;Qian Zhang ,&nbsp;Mahardhika Pratama","doi":"10.1016/j.knosys.2025.114514","DOIUrl":"10.1016/j.knosys.2025.114514","url":null,"abstract":"<div><div>Modern autonomous driving systems rely heavily on deep learning-based perception models for object detection; yet, their computational and energy demands remain critical bottlenecks. The existing adaptive-perception strategies often lack the ability to dynamically balance the detection accuracy and energy consumption, in real-time, particularly under varying environmental conditions. To address this challenge, we first construct a large-scale autonomous driving dataset based on the CARLA simulator. Then, we propose a novel metric—the balanced efficiency index—to annotate each image with the most suitable you-only-look-once version 8 (YOLOv8) model size (i.e., n, s, m, l, or x). This index is governed by two critical parameters, which are efficiently optimized using our proposed constrained stochastic DIviding RECTangles (DIRECT) algorithm. Finally, we propose a lightweight dynamic mixed receptive field transformer (DynaMixFormer), which is trained using the labelled dataset, to select the appropriate YOLOv8 model adaptively. Our results show that: (1) the constrained stochastic DIRECT algorithm determines cost-effective parameters with very limited simulation overhead; (2) DynaMixFormer achieves a high classification accuracy of 96.56 % with only 0.017 M parameters, outperforming the state-of-the-art image-classification networks; and (3) the well-trained DynaMixFormer effectively extracts real-time contextual features, such as traffic density, weather conditions, and road complexity, to intelligently select the optimal model from various YOLOv8 variants. Extensive simulations demonstrate that our approach achieves up to 70.20 % reduction in the energy consumption, compared to the static deployment of the YOLOv8x model, with only a marginal decrease of approximately 2 % in the mean average precision. Taking China as an example, this translates to an estimated energy saving of 2.73 × 10¹⁴ W. This work not only advances energy-efficient autonomous perception but also provides a generalizable framework for adaptive model selection in resource-constrained edge-computing systems. For ease of comprehension, some key nomenclature used in this paper are summarized in Table 1.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114514"},"PeriodicalIF":7.6,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159951","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":"Secure distributor data storage and retrieval of unstructured data in blockchain enabled edge computing","authors":"S. Premkumar ,&nbsp;S. Sivakumar ,&nbsp;TS. Arthi ,&nbsp;N. Partheeban","doi":"10.1016/j.knosys.2025.114518","DOIUrl":"10.1016/j.knosys.2025.114518","url":null,"abstract":"<div><div>In modern IoT applications, managing large volumes of unstructured data securely and efficiently is a growing challenge, especially within blockchain-enabled edge computing environments. Traditional data storage and retrieval methods often fall short in terms of error detection, indexing efficiency, and secure data handling. To address these limitations, this research proposes a secure and intelligent distributor framework for the storage and retrieval of unstructured data using a blockchain-supported edge computing model. The system architecture is composed of three layers, such as the IoT network layer for data collection, the blockchain-based edge computing layer for secure data handling, and the cloud layer for scalable storage. The proposed framework introduces a novel indexing mechanism, the Optimal Cluster Inverted Index (OCII), which is computed using a newly designed Taylor Fire Hawk Optimizer (Taylor FHO), which is the hybridization of the Taylor series and Fire Hawk Optimizer (FHO). The data handling framework involves five key processes, like KeyGeneration, OCII Generation, AuthGen, Check, and Dynamics, ensuring secure indexing, authentication, and data validation. Experimental evaluation demonstrates that the Taylor FHO achieves a better precision of 86.067%, recall of 87.080%, F-measure of 87.748%, and indexing time of 0.401 sec. This research provides a scalable and secure solution for real-time unstructured data processing in IoT systems.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114518"},"PeriodicalIF":7.6,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222225","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":"Data-driven interpretation of dimensions in an embedding language model based on a reference knowledge graph","authors":"Jose L. Mellina-Andreu ,&nbsp;Alejandro Cisterna-García ,&nbsp;Juan A. Botía","doi":"10.1016/j.knosys.2025.114507","DOIUrl":"10.1016/j.knosys.2025.114507","url":null,"abstract":"<div><div>As language models are used in more applications, a key problem has become clear: their numerical embeddings are hard to interpret because it is unclear how each part of the vector relates to real-world meanings in specific fields. The prevailing embedding methods are inadequate in their current state, as they are unable to effectively bridge the gap between mathematical representations and human-understandable knowledge structures. The present study proposes a novel framework that explicitly links ontology classes to specific embedding dimensions through a dual-component architecture combining a text encoder that produces the target embedding dimensions with domain knowledge graphs. The Area Under the Interpretability Curve (AUIC) metric is introduced as a means to systematically evaluate model-alignment with ontological concepts. The analysis reveals that targeted dimensional mapping enables direct interpretation of individual vector components through ontological terms. The practical applications of this framework are illustrated through case studies in biomedical contexts, demonstrating enhanced model transparency without compromising performance. This approach establishes a measurable pathway for reconciling statistical language representations with structured domain knowledge, particularly benefiting fields requiring precise concept alignment like biomedicine. The implementation is publicly available at: <span><span>https://github.com/Mellandd/DEIBO</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114507"},"PeriodicalIF":7.6,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159948","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 novel deep cognitive network for battlefield situation awareness in wargaming","authors":"Chenhui Pan ,&nbsp;Yong Xian ,&nbsp;Peiyang Ma ,&nbsp;Leliang Ren ,&nbsp;Wancheng Ni","doi":"10.1016/j.knosys.2025.114516","DOIUrl":"10.1016/j.knosys.2025.114516","url":null,"abstract":"<div><div>As an innovative approach to supporting wargaming, computer-based wargames have been well received by military researchers. The battlefield situation in wargames is complex and rapidly evolving, and analysing a single scenario is insufficient to capture the full scope of the battlefield. To address the challenge of identifying trends in situational changes, this study proposes a value network model for battlefield situational awareness in wargaming based on deep learning techniques. Focusing on the Army Tactical Wargame as the research object, this study analyses key elements of battlefield situations using feature engineering methods. It introduces a hierarchical, grid-based model for representing battlefield situation features within wargames and develops a value tagging system that integrates system scores with distance-based rewards. A convolutional neural network-based value network model for situational awareness is then constructed, and the influence of key battlefield characteristics on the model is examined. Experimental results demonstrate that the proposed value network can more accurately predict the situation value at each stage of the wargame. The prediction accuracy exhibits a hump-shaped trend from the beginning to the end of the simulation. During the attack phase, the prediction accuracy exceeds 70 %, reaching a peak of 72.98 %. These findings offer a reliable new method for supporting agents in situation recognition and intelligent decision-making.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114516"},"PeriodicalIF":7.6,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222224","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":"Adaptive exploration for few-shot incremental learning","authors":"Cao Han ,&nbsp;Ziqi Gu ,&nbsp;Chunyan Xu ,&nbsp;Zhen Cui","doi":"10.1016/j.knosys.2025.114496","DOIUrl":"10.1016/j.knosys.2025.114496","url":null,"abstract":"<div><div>Few-shot class incremental learning (FSCIL) poses a challenging problem in computer vision, where conventional deep models suffer from catastrophic forgetting and overfitting to novel classes. Inspired by the dynamic learning processes observed in human cognition when adapting to unfamiliar scenarios, we propose a deep exploratory incremental learning framework that incrementally refines the classifier model through a trial-and-error decision making process. A joint distribution-aware reward function is introduced to guide learning, incorporating three key factors: intra-class compactness, inter-class dispersion, and cross-session consistency, enabling balanced knowledge retention and acquisition. Furthermore, we design a dynamic gradient guidance module that adaptively adjusts gradient updates within a Gaussian-derived policy space, enhancing training stability and mitigating overfitting risks in the few shot regime. Extensive experiments conducted on three publicly available datasets demonstrate the effectiveness of the proposed method, achieving state-of-the-art performance in the FSCIL setting.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"330 ","pages":"Article 114496"},"PeriodicalIF":7.6,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159947","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}