Neurocomputing最新文献

英文中文

MoSCE-ReID: Mixture of semantic clustering experts for person re-identification

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129587

Kai Ren , Chuanping Hu , Hao Xi , Yongqiang Li , Jinhao Fan , Lihua Liu

This study advances the utilization of semantic information in person re-identification (ReID) by leveraging pre-trained vision-language models, addressing the current limitations in semantic information processing within ReID systems. While recent studies have explored CLIP integration for ReID tasks, their training approaches have inadvertently diminished semantic information by focusing primarily on indirect alignment between person IDs through text encoders and image features. Through comprehensive empirical analysis of semantic information’s role in pedestrian ReID, we propose MoSCE-ReID, a mixed semantic clustering expert model. The framework incorporates two key components: a learnable Attribute Group Weight Extractor (AGWE) and a Mixed of LoRA Expert (MoLE) module, designed specifically for attribute group feature extraction. The final ReID decisions are made through the synergistic integration of attribute group features and global features. Extensive experiments across multiple public datasets demonstrate that our approach, by effectively incorporating person attribute group semantic information, achieves substantial performance improvements in ReID tasks, exhibiting superior generalization capabilities compared to existing frameworks.

{"title":"MoSCE-ReID: Mixture of semantic clustering experts for person re-identification","authors":"Kai Ren , Chuanping Hu , Hao Xi , Yongqiang Li , Jinhao Fan , Lihua Liu","doi":"10.1016/j.neucom.2025.129587","DOIUrl":"10.1016/j.neucom.2025.129587","url":null,"abstract":"<div><div>This study advances the utilization of semantic information in person re-identification (ReID) by leveraging pre-trained vision-language models, addressing the current limitations in semantic information processing within ReID systems. While recent studies have explored CLIP integration for ReID tasks, their training approaches have inadvertently diminished semantic information by focusing primarily on indirect alignment between person IDs through text encoders and image features. Through comprehensive empirical analysis of semantic information’s role in pedestrian ReID, we propose MoSCE-ReID, a mixed semantic clustering expert model. The framework incorporates two key components: a learnable Attribute Group Weight Extractor (AGWE) and a Mixed of LoRA Expert (MoLE) module, designed specifically for attribute group feature extraction. The final ReID decisions are made through the synergistic integration of attribute group features and global features. Extensive experiments across multiple public datasets demonstrate that our approach, by effectively incorporating person attribute group semantic information, achieves substantial performance improvements in ReID tasks, exhibiting superior generalization capabilities compared to existing frameworks.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129587"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143314460","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}

引用次数: 0

LEFuse: Joint low-light enhancement and image fusion for nighttime infrared and visible images

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129592

Muhang Cheng , Haiyan Huang , Xiangyu Liu , Hongwei Mo , Xiongbo Zhao , Songling Wu

Infrared and visible image fusion (IVIF) aims to represent scenes more richly and accurately by integrating information from both modalities. However, existing IVIF methods are typically designed for normal illumination conditions, aiming to achieve higher scores by maintaining close similarity to the source images. In night scenes, visible images often suffer from both low light and localized overexposure due to the dim environment and the interference from local light sources. These methods fail to explore the information hidden in the dark regions of visible images, resulting in fusion images that lack texture details, appear overall dark, and exhibit poor visual quality. To address this issue, we propose a novel image fusion network called LEFuse. LEFuse not only integrates complementary information from both visible and infrared images but also focuses on recovering hidden texture details in visible images. By doing so, LEFuse enhances the visibility and contrast of the fused image, resulting in a brighter and more vivid representation. To achieve this goal, we propose a set of unsupervised loss functions to drive the network’s learning. This set includes a maximum entropy-based fusion enhancement loss for both image fusion and low-light enhancement, as well as a perceptual loss to mitigate the impact of local overexposure in visible images on the fused result. These losses can be applied to any existing image fusion network, enhancing fused images without compromising fusion performance. Extensive experiments demonstrate that our LEFuse achieves promising results in terms of visual quality and quantitative evaluations, especially in nighttime environments. Our code is publicly available at https://github.com/cheng411523/LEFuse.

{"title":"LEFuse: Joint low-light enhancement and image fusion for nighttime infrared and visible images","authors":"Muhang Cheng , Haiyan Huang , Xiangyu Liu , Hongwei Mo , Xiongbo Zhao , Songling Wu","doi":"10.1016/j.neucom.2025.129592","DOIUrl":"10.1016/j.neucom.2025.129592","url":null,"abstract":"<div><div>Infrared and visible image fusion (IVIF) aims to represent scenes more richly and accurately by integrating information from both modalities. However, existing IVIF methods are typically designed for normal illumination conditions, aiming to achieve higher scores by maintaining close similarity to the source images. In night scenes, visible images often suffer from both low light and localized overexposure due to the dim environment and the interference from local light sources. These methods fail to explore the information hidden in the dark regions of visible images, resulting in fusion images that lack texture details, appear overall dark, and exhibit poor visual quality. To address this issue, we propose a novel image fusion network called LEFuse. LEFuse not only integrates complementary information from both visible and infrared images but also focuses on recovering hidden texture details in visible images. By doing so, LEFuse enhances the visibility and contrast of the fused image, resulting in a brighter and more vivid representation. To achieve this goal, we propose a set of unsupervised loss functions to drive the network’s learning. This set includes a maximum entropy-based fusion enhancement loss for both image fusion and low-light enhancement, as well as a perceptual loss to mitigate the impact of local overexposure in visible images on the fused result. These losses can be applied to any existing image fusion network, enhancing fused images without compromising fusion performance. Extensive experiments demonstrate that our LEFuse achieves promising results in terms of visual quality and quantitative evaluations, especially in nighttime environments. Our code is publicly available at <span><span>https://github.com/cheng411523/LEFuse</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129592"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143378187","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}

引用次数: 0

Incremental value iteration for optimal output regulation of linear systems with unknown exosystems

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129579

Chonglin Jing , Chaoli Wang , Dong Liang , Yujing Xu , Longyan Hao

This paper addresses the optimal output regulation problem for discrete-time linear systems with completely unknown dynamics and unmeasurable exosystem states. The primary objective is to design incremental dataset-based value iteration (VI) reinforcement learning algorithms to derive both state feedback and output feedback controllers. In the context of data-driven optimal control, existing approaches typically require either the exosystem state to be measurable or the design of an autonomous system to reconstruct it. In contrast, this work proposes an incremental dataset-based VI algorithm, which eliminates the need for exosystem state measurement or reconstruction. Additionally, the proposed method allows for the selection of an arbitrary initial admissible control policy, thereby overcoming the challenge of requiring an initial admissible control in policy iteration algorithms. Furthermore, the system state is reconstructed using the incremental dataset, and an optimal output feedback controller is developed based on the proposed VI algorithm. The theoretical convergence of the dataset-based incremental VI algorithm is rigorously analyzed, and comprehensive simulations are conducted to validate its effectiveness.

{"title":"Incremental value iteration for optimal output regulation of linear systems with unknown exosystems","authors":"Chonglin Jing , Chaoli Wang , Dong Liang , Yujing Xu , Longyan Hao","doi":"10.1016/j.neucom.2025.129579","DOIUrl":"10.1016/j.neucom.2025.129579","url":null,"abstract":"<div><div>This paper addresses the optimal output regulation problem for discrete-time linear systems with completely unknown dynamics and unmeasurable exosystem states. The primary objective is to design incremental dataset-based value iteration (VI) reinforcement learning algorithms to derive both state feedback and output feedback controllers. In the context of data-driven optimal control, existing approaches typically require either the exosystem state to be measurable or the design of an autonomous system to reconstruct it. In contrast, this work proposes an incremental dataset-based VI algorithm, which eliminates the need for exosystem state measurement or reconstruction. Additionally, the proposed method allows for the selection of an arbitrary initial admissible control policy, thereby overcoming the challenge of requiring an initial admissible control in policy iteration algorithms. Furthermore, the system state is reconstructed using the incremental dataset, and an optimal output feedback controller is developed based on the proposed VI algorithm. The theoretical convergence of the dataset-based incremental VI algorithm is rigorously analyzed, and comprehensive simulations are conducted to validate its effectiveness.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129579"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143348774","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}

引用次数: 0

A survey of artificial intelligence in gait-based neurodegenerative disease diagnosis

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129533

Haocong Rao , Minlin Zeng , Xuejiao Zhao, Chunyan Miao

Recent years have witnessed an increasing global population affected by neurodegenerative diseases (NDs), which traditionally require extensive healthcare resources and human effort for medical diagnosis and monitoring. As a crucial disease-related motor symptom, human gait can be exploited to characterize different NDs. The current advances in artificial intelligence (AI) models enable automatic gait analysis for NDs identification and classification, opening a new avenue to facilitate faster and more cost-effective diagnosis of NDs. In this paper, we provide a comprehensive survey on recent progress of machine learning and deep learning based AI techniques applied to diagnosis of five typical NDs through gait. We provide an overview of the process of AI-assisted NDs diagnosis, and present a systematic taxonomy of existing gait data and AI models. Meanwhile, a novel quality evaluation criterion is proposed to quantitatively assess the quality of existing studies. Through an extensive review and analysis of 169 studies, we present recent technical advancements, discuss existing challenges, potential solutions, and future directions in this field. Finally, we envision the prospective utilization of 3D skeleton data for human gait representation and the development of more efficient AI models for NDs diagnosis. ²

{"title":"A survey of artificial intelligence in gait-based neurodegenerative disease diagnosis","authors":"Haocong Rao , Minlin Zeng , Xuejiao Zhao, Chunyan Miao","doi":"10.1016/j.neucom.2025.129533","DOIUrl":"10.1016/j.neucom.2025.129533","url":null,"abstract":"<div><div>Recent years have witnessed an increasing global population affected by neurodegenerative diseases (NDs), which traditionally require extensive healthcare resources and human effort for medical diagnosis and monitoring. As a crucial disease-related motor symptom, human gait can be exploited to characterize different NDs. The current advances in artificial intelligence (AI) models enable automatic gait analysis for NDs identification and classification, opening a new avenue to facilitate faster and more cost-effective diagnosis of NDs. In this paper, we provide a comprehensive survey on recent progress of machine learning and deep learning based AI techniques applied to diagnosis of five typical NDs through gait. We provide an overview of the process of AI-assisted NDs diagnosis, and present a systematic taxonomy of existing gait data and AI models. Meanwhile, a novel quality evaluation criterion is proposed to quantitatively assess the quality of existing studies. Through an extensive review and analysis of 169 studies, we present recent technical advancements, discuss existing challenges, potential solutions, and future directions in this field. Finally, we envision the prospective utilization of 3D skeleton data for human gait representation and the development of more efficient AI models for NDs diagnosis. <span><span><sup>2</sup></span></span></div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129533"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143355503","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}

引用次数: 0

Adaptive node similarity for DropEdge

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129574

Yangcai Xie, Jiecheng Li, Shichao Zhang

There are two principal impediments of expanding deep graph convolutional networks (GCNs) due to the assumption of smoothness, i.e., over-fitting and over-smoothing. DropEdge methods relieve the convergence speed and reduce the information loss by randomly dropping a specific rate of edges, hence effectively alleviates these two issues and has been widely used to many backbone models. However, thanks to the blindness and potential risks of randomly removing edges, current DropEdge methods often remove important edges and retain unimportant edges, this inevitably reduce the accuracy of learning results. In order to tackle the challenges in previous DropEdge methods, this paper proposes a precise removal technique through the node similarity, which is closely related to edges. Specifically, we employ the hybrid optimal node similarity to drop edges, on the one hand, the edges that severely affect over-fitting and over-smoothing of nodes, i.e., the edges with high node similarity, are removed; on the other hand, the edges that are outliers and noisy, i.e., the edges with a large difference in similarity to normal nodes, are also removed. Therefore, our methods significantly alleviate over-fitting and over-smoothing, accurately reduce the impact of outliers and noise, more importantly, our methods is a generic skill that can be deployed current GCN and its variants. Experimental results on seven benchmark datasets including three assortative datasets and four disassortative datasets show that our methods outperforms the state-of-the-art methods, improve the performance by a large margin especially for disassortative graphs.

{"title":"Adaptive node similarity for DropEdge","authors":"Yangcai Xie, Jiecheng Li, Shichao Zhang","doi":"10.1016/j.neucom.2025.129574","DOIUrl":"10.1016/j.neucom.2025.129574","url":null,"abstract":"<div><div>There are two principal impediments of expanding deep graph convolutional networks (GCNs) due to the assumption of smoothness, <em>i.e.,</em> over-fitting and over-smoothing. DropEdge methods relieve the convergence speed and reduce the information loss by randomly dropping a specific rate of edges, hence effectively alleviates these two issues and has been widely used to many backbone models. However, thanks to the blindness and potential risks of randomly removing edges, current DropEdge methods often remove important edges and retain unimportant edges, this inevitably reduce the accuracy of learning results. In order to tackle the challenges in previous DropEdge methods, this paper proposes a precise removal technique through the node similarity, which is closely related to edges. Specifically, we employ the hybrid optimal node similarity to drop edges, on the one hand, the edges that severely affect over-fitting and over-smoothing of nodes, i.e., the edges with high node similarity, are removed; on the other hand, the edges that are outliers and noisy, i.e., the edges with a large difference in similarity to normal nodes, are also removed. Therefore, our methods significantly alleviate over-fitting and over-smoothing, accurately reduce the impact of outliers and noise, more importantly, our methods is a generic skill that can be deployed current GCN and its variants. Experimental results on seven benchmark datasets including three assortative datasets and four disassortative datasets show that our methods outperforms the state-of-the-art methods, improve the performance by a large margin especially for disassortative graphs.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129574"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143348623","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}

引用次数: 0

Unified finite-time error analysis of soft Q-learning

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129582

Narim Jeong, Donghwan Lee

Soft Q-learning is one of the most commonly used reinforcement learning algorithms for various purposes, e.g., dealing with entropy-regularized Markov decision problems, reducing the overestimation bias, and improving explorations. Its effectiveness in practice has led to its widespread use; however, there has not been much theoretical study on soft Q-learning. This paper attempts to provide an integrated finite-time analytical approach for soft Q-learning from a control-theoretic perspective. We examine three different kinds of soft Q-learning algorithms that use the log-sum-exp operator, the Boltzmann operator, and the mellowmax operator, respectively. Utilizing dynamical switching system models, we obtain the finite-time error bounds of three soft Q-learning variants. We believe that our analysis can assist in a better understanding of soft Q-learning through links with switching system models.

引用次数: 0

Stability analysis of inertial delayed neural network with delayed impulses via dynamic event-triggered impulsive control

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129573

Mengyao Shi , Lulu Li , Jinde Cao , Liang Hua , Mahmoud Abdel-Aty

This paper investigates the stability of inertial delayed neural network under dynamic event-triggered impulsive control (DETIC). We innovate by generating the impulsive sequence through DETIC and incorporating impulsive delays, thereby enhancing the model’s practical relevance. Our methodology involves a two-step process: first, we transform the inertial neural network into a first-order differential form using appropriate vector transformations. Then, leveraging Lyapunov-based dynamic event-triggered control, we derive sufficient conditions for both uniform stability and uniform asymptotic stability of the system. To ensure practical applicability, we establish specific parameter constraints for the DETIC mechanism that precludes the Zeno phenomenon. To demonstrate the accuracy and efficacy of our theoretical results, we present two simulation examples.

引用次数: 0

Privacy-preserving face attribute classification via differential privacy

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129556

Xiaoting Zhang , Tao Wang , Junhao Ji , Yushu Zhang , Rushi Lan

The development of face attribute recognition technology has enhanced the intelligence capabilities in the retail industry. Merchants use the surveillance system to capture customers’ face images, and analyze their basic characteristics to provide accurate product recommendations and optimize product configurations. However, these captured face images may contain sensitive visual information, especially identity-related data, which could lead to potential security and privacy risks. Current methods for face privacy protection cannot fully support privacy preserving face attributes classification. To this end, this paper proposes a privacy protection scheme that employs differential privacy in the frequency domain to mitigate risks in face attribute classification systems. Our main goal is to take the frequency domain features perturbed with differential privacy as the input of the face attribute classification model to resist privacy attacks. Specifically, the proposed scheme first transforms the original face image into the frequency domain using the discrete cosine transform (DCT) and removes the DC components that contain the visual information. Then the privacy budget allocation in the differential privacy framework is optimized based on the loss of the face attribute classification network. Finally, the corresponding differential privacy noise is added to the frequency representation. The utilization of differential privacy theoretically provides privacy guarantees. Sufficient experimental results show that the proposed scheme can well balance the privacy-utility.

{"title":"Privacy-preserving face attribute classification via differential privacy","authors":"Xiaoting Zhang , Tao Wang , Junhao Ji , Yushu Zhang , Rushi Lan","doi":"10.1016/j.neucom.2025.129556","DOIUrl":"10.1016/j.neucom.2025.129556","url":null,"abstract":"<div><div>The development of face attribute recognition technology has enhanced the intelligence capabilities in the retail industry. Merchants use the surveillance system to capture customers’ face images, and analyze their basic characteristics to provide accurate product recommendations and optimize product configurations. However, these captured face images may contain sensitive visual information, especially identity-related data, which could lead to potential security and privacy risks. Current methods for face privacy protection cannot fully support privacy preserving face attributes classification. To this end, this paper proposes a privacy protection scheme that employs differential privacy in the frequency domain to mitigate risks in face attribute classification systems. Our main goal is to take the frequency domain features perturbed with differential privacy as the input of the face attribute classification model to resist privacy attacks. Specifically, the proposed scheme first transforms the original face image into the frequency domain using the discrete cosine transform (DCT) and removes the DC components that contain the visual information. Then the privacy budget allocation in the differential privacy framework is optimized based on the loss of the face attribute classification network. Finally, the corresponding differential privacy noise is added to the frequency representation. The utilization of differential privacy theoretically provides privacy guarantees. Sufficient experimental results show that the proposed scheme can well balance the privacy-utility.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129556"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143314457","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}

引用次数: 0

A contrastive learning strategy for optimizing node non-alignment in dynamic community detection

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129548

Xiaohong Li, Wanyao Shi, Qixuan Peng, Hongyan Ran

Dynamic community detection, which focuses on tracking local topological variation with time, is crucial for understanding the changing affiliations of nodes to communities in complex networks. Existing researches fell short of expectations primarily due to their heavy reliance on clustering methods or evolutionary algorithms. The emergence of graph contrastive learning offers us a novel perspective and inspiration, which performed well in recognizing pattern at both the node-node and node-graph levels. However, there are still the following limitations in practice: (i) conventional data augmentations may undermine task-relevant information by bring in invalid views or false positive samples, leading the model toward weak discriminative representations. (ii) the non-alignment of nodes caused by dynamic changes also limits the expressive ability of GCL. In this paper, we propose a Contrastive Learning strategy for Optimizing Node non-alignment in Dynamic Community Detection (CL-OND). Initially, we confirm the viability of utilizing dynamic adjacent snapshots as monitoring signals through graph spectral experiments, which eliminates the dependence of contrastive learning on traditional data augmentations. Subsequently, we construct an end-to-end dynamic community detection model and introduce a non-aligned neighbor contrastive loss to capture temporal properties and inherent structure of evolutionary graphs by constructing positive and negative samples. Furthermore, extensive experimental results demonstrate that our approach consistently outperforms others in terms of performance.

{"title":"A contrastive learning strategy for optimizing node non-alignment in dynamic community detection","authors":"Xiaohong Li, Wanyao Shi, Qixuan Peng, Hongyan Ran","doi":"10.1016/j.neucom.2025.129548","DOIUrl":"10.1016/j.neucom.2025.129548","url":null,"abstract":"<div><div>Dynamic community detection, which focuses on tracking local topological variation with time, is crucial for understanding the changing affiliations of nodes to communities in complex networks. Existing researches fell short of expectations primarily due to their heavy reliance on clustering methods or evolutionary algorithms. The emergence of graph contrastive learning offers us a novel perspective and inspiration, which performed well in recognizing pattern at both the node-node and node-graph levels. However, there are still the following limitations in practice: (i) conventional data augmentations may undermine task-relevant information by bring in invalid views or false positive samples, leading the model toward weak discriminative representations. (ii) the non-alignment of nodes caused by dynamic changes also limits the expressive ability of GCL. In this paper, we propose a <strong>C</strong>ontrastive <strong>L</strong>earning strategy for <strong>O</strong>ptimizing <strong>N</strong>ode non-alignment in <strong>D</strong>ynamic Community Detection (<strong>CL-OND</strong>). Initially, we confirm the viability of utilizing dynamic adjacent snapshots as monitoring signals through graph spectral experiments, which eliminates the dependence of contrastive learning on traditional data augmentations. Subsequently, we construct an end-to-end dynamic community detection model and introduce a non-aligned neighbor contrastive loss to capture temporal properties and inherent structure of evolutionary graphs by constructing positive and negative samples. Furthermore, extensive experimental results demonstrate that our approach consistently outperforms others in terms of performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"626 ","pages":"Article 129548"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143348779","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}

引用次数: 0

DKETFormer: Salient object detection in optical remote sensing images based on discriminative knowledge extraction and transfer

IF 5.5 2区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Neurocomputing

Pub Date : 2025-02-04 DOI: 10.1016/j.neucom.2025.129558

Yuze Sun, Hongwei Zhao, Jianhang Zhou

Generally, most methods for salient object detection in optical remote sensing images (ORSI-SOD) are based on convolutional neural networks (CNNs). However, CNNs, due to their architectural characteristics, can only encode local semantic information, which leads to a lack of exploration of discriminative features on a large scale. Therefore, to encode the long-term dependency within the detection image, enhance the extraction of discriminative knowledge, and transfer it at multiple scales, we introduce a Transformer architecture called DKETFormer. Specifically, DKETFormer utilizes the Transformer backbone to obtain multi-scale feature maps that have encoded long-term dependency relationships. Then, it constructs a decoder using the Cross-spatial Knowledge Extraction Module (CKEM) and the Inter-layer Feature Transfer Module (IFTM). The CKEM is capable of extracting discriminative information across receptive fields while preserving knowledge from each channel. It also utilizes global information encoding to calibrate channel weights, resulting in improved knowledge aggregation and capturing of pixel-level pairwise relationships. The IFTM utilizes encoded and extracted information from the backbone and CKEM, employing a self-attention mechanism with cosine similarity knowledge to model and propagate discriminative features. Finally, we generated the final detection map using a salient object detector. The results of comparative experiments and ablation experiments demonstrate the effectiveness of the proposed DKETFormer and its internal modules.

{"title":"DKETFormer: Salient object detection in optical remote sensing images based on discriminative knowledge extraction and transfer","authors":"Yuze Sun, Hongwei Zhao, Jianhang Zhou","doi":"10.1016/j.neucom.2025.129558","DOIUrl":"10.1016/j.neucom.2025.129558","url":null,"abstract":"<div><div>Generally, most methods for salient object detection in optical remote sensing images (ORSI-SOD) are based on convolutional neural networks (CNNs). However, CNNs, due to their architectural characteristics, can only encode local semantic information, which leads to a lack of exploration of discriminative features on a large scale. Therefore, to encode the long-term dependency within the detection image, enhance the extraction of discriminative knowledge, and transfer it at multiple scales, we introduce a Transformer architecture called DKETFormer. Specifically, DKETFormer utilizes the Transformer backbone to obtain multi-scale feature maps that have encoded long-term dependency relationships. Then, it constructs a decoder using the Cross-spatial Knowledge Extraction Module (CKEM) and the Inter-layer Feature Transfer Module (IFTM). The CKEM is capable of extracting discriminative information across receptive fields while preserving knowledge from each channel. It also utilizes global information encoding to calibrate channel weights, resulting in improved knowledge aggregation and capturing of pixel-level pairwise relationships. The IFTM utilizes encoded and extracted information from the backbone and CKEM, employing a self-attention mechanism with cosine similarity knowledge to model and propagate discriminative features. Finally, we generated the final detection map using a salient object detector. The results of comparative experiments and ablation experiments demonstrate the effectiveness of the proposed DKETFormer and its internal modules.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"625 ","pages":"Article 129558"},"PeriodicalIF":5.5,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143210303","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}

引用次数: 0

首页上一页

下一页尾页

类型

全部化学•材料生命科学医学物理工程技术环境•农林材料科学地球科学法学管理学化学环境科学与生态学计算机科学教育学经济学农林科学人文科学生物学数学物理与天体物理心理学综合性期刊其他工业工程理学历史学农学文学信息工程

数据库

全部 ACS Publications Elsevier ieeexplore Springer The Royal Society of Chemistry Wiley

期刊

Neurocomputing

全部 Acc. Chem. Res. ACS Applied Bio Materials ACS Appl. Electron. Mater. ACS Appl. Energy Mater. ACS Appl. Mater. Interfaces ACS Appl. Nano Mater. ACS Appl. Polym. Mater. ACS BIOMATER-SCI ENG ACS Catal. ACS Cent. Sci. ACS Chem. Biol. ACS Chemical Health & Safety ACS Chem. Neurosci. ACS Comb. Sci. ACS Earth Space Chem. ACS Energy Lett. ACS Infect. Dis. ACS Macro Lett. ACS Mater. Lett. ACS Med. Chem. Lett. ACS Nano ACS Omega ACS Photonics ACS Sens. ACS Sustainable Chem. Eng. ACS Synth. Biol. Anal. Chem. BIOCHEMISTRY-US Bioconjugate Chem. BIOMACROMOLECULES Chem. Res. Toxicol. Chem. Rev. Chem. Mater. CRYST GROWTH DES ENERG FUEL Environ. Sci. Technol. Environ. Sci. Technol. Lett. Eur. J. Inorg. Chem. IND ENG CHEM RES Inorg. Chem. J. Agric. Food. Chem. J. Chem. Eng. Data J. Chem. Educ. J. Chem. Inf. Model. J. Chem. Theory Comput. J. Med. Chem. J. Nat. Prod. J PROTEOME RES J. Am. Chem. Soc. LANGMUIR MACROMOLECULES Mol. Pharmaceutics Nano Lett. Org. Lett. ORG PROCESS RES DEV ORGANOMETALLICS J. Org. Chem. J. Phys. Chem. J. Phys. Chem. A J. Phys. Chem. B J. Phys. Chem. C J. Phys. Chem. Lett. Analyst Anal. Methods Biomater. Sci. Catal. Sci. Technol. Chem. Commun. Chem. Soc. Rev. CHEM EDUC RES PRACT CRYSTENGCOMM Dalton Trans. Energy Environ. Sci. ENVIRON SCI-NANO ENVIRON SCI-PROC IMP ENVIRON SCI-WAT RES Faraday Discuss. Food Funct. Green Chem. Inorg. Chem. Front. Integr. Biol. J. Anal. At. Spectrom. J. Mater. Chem. A J. Mater. Chem. B J. Mater. Chem. C Lab Chip Mater. Chem. Front. Mater. Horiz. MEDCHEMCOMM Metallomics Mol. Biosyst. Mol. Syst. Des. Eng. Nanoscale Nanoscale Horiz. Nat. Prod. Rep. New J. Chem. Org. Biomol. Chem. Org. Chem. Front. PHOTOCH PHOTOBIO SCI PCCP Polym. Chem.

﹀