计算机科学最新文献

{"title":"Inductive multiple clustering based on weakly-supervised salient representation learning","authors":"Wenjie Zhu ,&nbsp;Wei Qi Yan","doi":"10.1016/j.eswa.2025.129082","DOIUrl":"10.1016/j.eswa.2025.129082","url":null,"abstract":"<div><div>The increasing recognition of data diversity has highlighted multiple clustering as a valuable approach for generating diverse clustering solutions. However, conventional methods often prioritize non-redundant clusterings across disjoint subspaces, potentially overlooking key data characteristics and limiting interpretability. In this paper, we propose an Inductive Multiple Clustering (IMC) framework designed to extract distinct and interpretable representations through weakly-supervised learning from diverse clustering perspectives. Specifically, IMC decomposes data objects into group-specific salient components using reconstruction and transformation matrices with low-rank and sparse regularization. To enhance diversity among clusters, an incoherent regularization minimizes similarities between group-specific transformations in a weakly-supervised manner. Unlike previous approaches, our framework emphasizes salient representations and integrates inductive learning into multiple clustering, facilitating comprehensive interpretations of clustering results. We employ the Alternating Direction Method of Multipliers (ADMM) to optimize IMC, leveraging resulting matrices for clustering diverse datasets. Experimental results on benchmark datasets demonstrate IMC’s superiority over existing methods, providing a comprehensive explanation of multiple clustering results and successful extension to unseen data clustering.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"296 ","pages":"Article 129082"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704224","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":"Attention based network for real-time road drivable area, lane line detection and scene identification","authors":"Feng You ,&nbsp;Yi Xie ,&nbsp;Siyi Zhang ,&nbsp;Hao Chen ,&nbsp;Haiwei Wang ,&nbsp;Wei Zhang ,&nbsp;Jianrong Liu","doi":"10.1016/j.engappai.2025.111781","DOIUrl":"10.1016/j.engappai.2025.111781","url":null,"abstract":"<div><div>The detection of road drivable areas and lane lines is considered a fundamental component of autonomous driving systems. However, most existing approaches handle these tasks independently, and multi-task networks frequently neglect the inherent correlation between them while failing to differentiate various lane line types. In practice, the delineation of drivable regions is strongly influenced by both lane line characteristics and contextual street scenes. To address these limitations, a novel multi-task network—Real-time Road Drivable Area, Lane Line Detection, and Scene Identification Network (RLSNet)—is proposed. This network is designed to perform simultaneous segmentation of drivable areas, detection of lane lines, and classification of road scenes. Drivable area estimation is optimized through the integration of lane and scene cues, guided by traffic regulations. A Residual Network (ResNet)-based backbone is employed, enhanced with Bidirectional Fusion Attention (BFA) for feature encoding. This is followed by a decoder incorporating a Feature Aggregation Module (FAM) to enable effective semantic–spatial fusion. Lane line detection is further refined using a Bilateral Up-Sampling Decoder (BUSD), while scene understanding is enhanced via a Scene Classification Module (SCM). Extensive experiments conducted on the challenging Berkeley DeepDrive 100K(BDD100K) dataset have demonstrated that RLSNet achieves high accuracy in both drivable area and lane line detection by leveraging the mutual guidance of lane and scene information. Furthermore, the network maintains real-time inference speed at 93 frames per second (FPS), striking a practical balance between semantic fidelity and computational efficiency for real-world deployment. The implementation code has been made publicly available at: <span><span>https://github.com/033186ZSY/RLSNet-master</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"160 ","pages":"Article 111781"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711687","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Expert system for non-technical loss detection in power distribution grids using particle swarm optimization and nested power flow integration","authors":"Nikolaos-Antonios Livanos ,&nbsp;Nikolaos Giamarelos ,&nbsp;Alex Alexandridis ,&nbsp;Elias N. Zois","doi":"10.1016/j.eswa.2025.128997","DOIUrl":"10.1016/j.eswa.2025.128997","url":null,"abstract":"<div><div>Non-Technical Losses in power distribution grids, primarily caused by electricity theft and meter inaccuracies, pose a significant challenge to utility companies, impacting revenue and grid stability. This paper introduces a novel expert system integrating nested Power Flow analysis with swarm intelligence for accurate Non-Technical Losses detection and localization in distribution grids. The proposed system leverages smart meter data, grid topology, and substation transformer readings to formulate an optimization problem in which the Active Power and Reactive Power consumption of one or more consumers is estimated using Particle Swarm Optimization. A key feature of the system is its adaptability to various operational scenarios, such as grid size, topology uncertainty, and distributed energy resource penetration. Extensive experimental results confirm the effectiveness of the proposed NTL detection method. In simulated scenarios, the worst-case Mean Absolute Error for Active Power estimation is limited to 0.1391 kW, with a corresponding mean actual consumption of 1.004 kW. For Reactive Power, the Mean Absolute Error does not exceed 0.0175 kVAr, relative to a mean consumption of 0.1072 kVAr. When evaluated on real consumption data, the worst-case Mean Absolute Error for Active Power across all smart meters is 0.1028 kW, with a mean actual consumption of 5.1333 kW, while for Reactive Power the worst-case Mean Absolute Error reaches 0.1385 kVAr against a mean consumption of 2.3433 kVAr. Furthermore, in the specific case where real active and reactive consumption is zero the proposed method maintains Mean Absolute Error values of 0.8095 kW for Active power and 0.0944 kVAr for Reactive power, thus verifying its reliable performance in avoiding false positives.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"296 ","pages":"Article 128997"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704225","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 Graph-based State Representation Learning for episodic reinforcement learning in task-oriented dialogue systems","authors":"Yasaman Saffari,&nbsp;Javad Salimi Sartakhti","doi":"10.1016/j.engappai.2025.111793","DOIUrl":"10.1016/j.engappai.2025.111793","url":null,"abstract":"<div><div>Recent research in dialogue state tracking has made significant progress in tracking user goals using pretrained language models and context-driven approaches. However, existing work has primarily focused on contextual representations, often overlooking the structural complexity and topological properties of state transitions in episodic reinforcement learning tasks.</div><div>In this study, we introduce a cutting-edge, dual-perspective state representation approach that provides a dynamic and inductive method for topological state representation learning in episodic reinforcement learning within task-oriented dialogue systems. The proposed model extracts inherent topological information from state transitions in the Markov Decision Process graph by employing a modified clustering technique to address the limitations of transductive graph representation learning. It inductively captures structural relationships and enables generalization to unseen states.</div><div>Another key innovation of this approach is the incorporation of dynamic graph representation learning with task-specific rewards using Temporal Difference error. This captures topological features of state transitions, allowing the system to adapt to evolving goals and enhance decision-making in task-oriented dialogue systems.</div><div>Experiments, including ablation studies, comparisons with existing approaches, and interpretability analysis, reveal that the proposed model significantly outperforms traditional contextual state representations, improving task success rates by 9%–13% across multiple domains. It also surpasses state-of-the-art Q-network-based methods, enhancing adaptability and decision-making in domains such as movie-ticket booking, restaurant reservations, and taxi ordering.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"160 ","pages":"Article 111793"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711686","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A deep architecture for in-vehicle intrusion detection using controller area network-graph relied feature images","authors":"Sreelekshmi M.S.,&nbsp;Aji S.","doi":"10.1016/j.compeleceng.2025.110584","DOIUrl":"10.1016/j.compeleceng.2025.110584","url":null,"abstract":"<div><div>Active integration of digital innovations into automotive systems anticipates flawless security and safety in automation, and it is a major concern in intelligent transport systems. Securing the in-vehicle Controller Area Network (CAN) bus communication remains a major challenge due to inherent protocol vulnerabilities. In this study, we propose a novel deep learning-based intrusion detection approach that captures structural and contextual patterns within CAN traffic. Specifically, we introduce the concept of CANGraph-feature images, where CAN message interactions combining payload and arbitration ID information are represented as graph structures and subsequently transformed into images. This transformation enables the use of Convolutional Neural Networks (CNNs), leveraging their powerful spatial feature extraction capabilities to detect subtle and complex anomalies. Our optimized CNN architecture automatically learns discriminative features from the CANGraph images, effectively identifying abnormal behaviors. Extensive experiments on real-world vehicular datasets demonstrate that the proposed method robustly detects a wide range of attack types, including replay, fuzzing, DoS, and spoofing. The presented deep architecture shows promising potential to enhance the security of in-vehicle networks by achieving strong performance while maintaining low computational overhead and latency.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"127 ","pages":"Article 110584"},"PeriodicalIF":4.0,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704336","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Advancing automatic photovoltaic defect detection using semi-supervised semantic segmentation of electroluminescence images","authors":"Abhishek Jha ,&nbsp;Yogesh Rawat ,&nbsp;Shruti Vyas","doi":"10.1016/j.engappai.2025.111790","DOIUrl":"10.1016/j.engappai.2025.111790","url":null,"abstract":"<div><div>Photovoltaic (PV) systems allow us to tap into all abundant solar energy, however they require regular maintenance for high efficiency and to prevent degradation. Traditional manual health check, using Electroluminescence (EL) imaging, is expensive and logistically challenging which makes automated defect detection essential. Current automation approaches require extensive manual expert labeling, which is time-consuming, expensive, and prone to errors. We propose PV-S3 (<strong>P</strong>hoto<strong>v</strong>oltaic-<strong>S</strong>emi-supervised <strong>S</strong>emantic <strong>S</strong>egmentation), a Semi-Supervised Learning approach for semantic segmentation of defects in EL images that reduces reliance on extensive labeling. PV-S3 is an artificial intelligence (AI) model trained using a few labeled images along with numerous unlabeled images. We introduce a novel Semi Cross-Entropy loss function to deal with class imbalance. We evaluate PV-S3 on multiple datasets and demonstrate its effectiveness and adaptability. With merely 20% labeled samples, we achieve an absolute improvement of 9.7% in mean Intersection-over-Union (mIoU), 13.5% in Precision, 29.15% in Recall, and 20.42% in F1-Score over prior state-of-the-art supervised method (which uses 100% labeled samples) on University of Central Florida-Electroluminescence (UCF-EL) dataset (largest dataset available for semantic segmentation of EL images) showing improvement in performance while reducing the annotation costs by 80%. For more details, visit our GitHub repository: <span><span>https://github.com/abj247/PV-S3</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"160 ","pages":"Article 111790"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Finding the underlying viscoelastic constitutive equation via universal differential equations and differentiable physics","authors":"Elias C. Rodrigues ,&nbsp;Roney L. Thompson ,&nbsp;Dário A.B. Oliveira ,&nbsp;Roberto F. Ausas","doi":"10.1016/j.engappai.2025.111788","DOIUrl":"10.1016/j.engappai.2025.111788","url":null,"abstract":"<div><div>Determining the appropriate constitutive model to describe the behavior of a given material is a fundamental, yet challenging, aspect of rheology. While data-driven methods present a promising path for refining these models, a more in-depth investigation into the capabilities and limitations of emerging techniques is required. This research addresses this gap by employing Universal Differential Equations (UDEs) and differentiable physics to model viscoelastic fluids, merging conventional differential equations with neural networks to reconstruct missing terms in constitutive models. This study focuses on analyzing four viscoelastic models, Upper Convected Maxwell (UCM), Johnson–Segalman, Giesekus, and Exponential Phan–Thien–Tanner (ePTT) using synthetic datasets. The methodology was tested across different experimental conditions, including oscillatory and startup flows. Relative error analyses revealed that the UDEs framework maintains low and stable errors (below 0.3%) for the UCM, Johnson–Segalman, and Giesekus models under various conditions, while exhibiting higher but consistent errors (4%) for the ePTT model due to its strong nonlinearity. These findings highlight the potential of UDEs in fluid mechanics while also identifying critical areas for methodological improvement. Additionally, a model distillation approach was employed to extract simplified models from complex ones, emphasizing the versatility and robustness of UDEs in rheological modeling.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"160 ","pages":"Article 111788"},"PeriodicalIF":7.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711689","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Stream Signal Separation Based on Asynchronous Control Meta-Surface Antenna","authors":"Yuze Guo,&nbsp;Liang Jin,&nbsp;Yangming Lou,&nbsp;Xiaoming Xu,&nbsp;Qinlong Li,&nbsp;Boming Li,&nbsp;Shuaiyin Wang","doi":"10.1049/cmu2.70062","DOIUrl":"https://doi.org/10.1049/cmu2.70062","url":null,"abstract":"<p>The real-time reconfigurable characteristics of meta-surface antennas can be used to separate multi-stream signals under the condition of single radio frequency (RF). However, with the increase of the symbol rate and the number of antenna arrays in the future, it will face the problem that the state switch rate of the electromagnetic unit is not enough to reach the upper limit of array effective degrees of freedom (DOF) of the meta-surface antenna. To solve this problem, a theory of asynchronous control meta-surface antenna is proposed in this paper. By designing the starting time of different element state switching, different electromagnetic element states are staggered to improve the array effective DOF of the meta-surface antenna. Then, an electromagnetic unit state design algorithm of asynchronous control meta-surface antenna based on the minimum condition number of equivalent channel matrix is proposed. We improve the sparrow search algorithm to solve the condition number minimization problem in order to obtain the better multi-stream signal separation performance. The simulation results show that compared with the synchronous control meta-surface antenna, theory proposed in this paper can improve the effective DOF of array under the condition of limited switch rate, and can effectively reduce the receiving bit error rate and improve spectral efficiency when separating multi-stream signals.</p>","PeriodicalId":55001,"journal":{"name":"IET Communications","volume":"19 1","pages":""},"PeriodicalIF":1.5,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/cmu2.70062","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695814","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Mamba Adaptive Anomaly Transformer with association discrepancy for time series","authors":"Abdellah Zakaria Sellam ,&nbsp;Ilyes Benaissa ,&nbsp;Abdelmalik Taleb-Ahmed ,&nbsp;Luigi Patrono ,&nbsp;Cosimo Distante","doi":"10.1016/j.engappai.2025.111685","DOIUrl":"10.1016/j.engappai.2025.111685","url":null,"abstract":"<div><div>Anomaly detection in time series poses a critical challenge in industrial monitoring, environmental sensing, and infrastructure reliability, where accurately distinguishing anomalies from complex temporal patterns remains an open problem. While existing methods, such as the Anomaly Transformer leveraging multi-layer association discrepancy between prior and series distributions and Dual Attention Contrastive Representation Learning architecture (DCdetector) employing dual-attention contrastive learning, have advanced the field, critical limitations persist. These include sensitivity to short-term context windows, computational inefficiency, and degraded performance under noisy and non-stationary real-world conditions. To address these challenges, we present MAAT (Mamba Adaptive Anomaly Transformer), an enhanced architecture that refines association discrepancy modeling and reconstruction quality for more robust anomaly detection. Our work introduces two key contributions to the existing Anomaly transformer architecture: Sparse Attention, which computes association discrepancy more efficiently by selectively focusing on the most relevant time steps. This reduces computational redundancy while effectively capturing long-range dependencies critical for discerning subtle anomalies. A Mamba-Selective State Space Model (Mamba-SSM) is also integrated into the reconstruction module. A skip connection bridges the original reconstruction and the Mamba-SSM output, while a Gated Attention mechanism adaptively fuses features from both pathways. This design balances fidelity and contextual enhancement dynamically, improving anomaly localization and overall detection performance. Extensive experiments on benchmark datasets demonstrate that MAAT significantly outperforms prior methods, achieving superior anomaly distinguishability and generalization across diverse time series applications. By addressing the limitations of existing approaches, MAAT sets a new standard for unsupervised time series anomaly detection in real-world scenarios. Code available at <span><span>https://github.com/ilyesbenaissa/MAAT</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"160 ","pages":"Article 111685"},"PeriodicalIF":7.5,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702857","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Enhanced Hybrid Metaheuristic for Hierarchical Scheduling in 4WIDS Multi-robot Systems under Confined Environments","authors":"Lin Zhang,&nbsp;Yichen An,&nbsp;Tianwei Niu,&nbsp;Runjiao Bao,&nbsp;Shoukun Wang,&nbsp;Junzheng Wang","doi":"10.1016/j.conengprac.2025.106498","DOIUrl":"10.1016/j.conengprac.2025.106498","url":null,"abstract":"<div><div>Multi-robot systems have emerged as a transformative paradigm for industrial automation. However, deploying these systems in dense, dynamic environments like ultra-dense warehouses and Ro-Ro terminals remains challenging due to simplified motion constraints, idealized models, and the tight coupling of task assignment, trajectory planning, and conflict resolution under strict spatiotemporal constraints. To address these problems, we propose a hierarchical scheduling framework for four-wheel independent drive/steering robot groups in confined environments. Firstly, at the task assignment layer, we introduce an enhanced hybrid metaheuristic for task assignment that integrates particle swarm optimization with a genetic algorithm, augmented by a problem-specific fitness function and adaptive mutation strategies to prevent premature convergence. Secondly, at the path planning layer, we develop a kinematics-aware conflict-based search path planner integrating motion primitives with improved A* node expansion strategies, where adaptive heuristic weighting and bidirectional search acceleration are introduced to ensure computational tractability. Simulations in a near-realistic confined environment show that the proposed hierarchical scheduling algorithm reduces total execution cost by 11.0% compared to the advanced particle swarm genetic algorithm, demonstrating its superior performance in multi-robot coordination. Furthermore, field tests conducted at the Ro-Ro Terminal of Yantai Port have fully validated the feasibility of this framework for multi-robot coordination in real-world scenarios. This work lays a theoretical and practical foundation for next-generation multi-robot coordination in constrained logistics ecosystems.</div></div>","PeriodicalId":50615,"journal":{"name":"Control Engineering Practice","volume":"164 ","pages":"Article 106498"},"PeriodicalIF":5.4,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144703675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}