NeurocomputingPub Date : 2025-03-20DOI: 10.1016/j.neucom.2025.130008
Yilin Wang , Sha Zhao , Haiteng Jiang , Shijian Li , Tao Li , Gang Pan
{"title":"M-MDD: A multi-task deep learning framework for major depressive disorder diagnosis using EEG","authors":"Yilin Wang , Sha Zhao , Haiteng Jiang , Shijian Li , Tao Li , Gang Pan","doi":"10.1016/j.neucom.2025.130008","DOIUrl":"10.1016/j.neucom.2025.130008","url":null,"abstract":"<div><div>Major depressive disorder (MDD) is a common and destructive psychiatric disorder worldwide. Traditional MDD diagnosis relies heavily on subjective observation and questionnaires. Recently, a non-invasive method of recording the brain’s spontaneous activity called Electroencephalogram (EEG) has been a useful tool of MDD diagnosis. However, there are still some challenges to be addressed: (1) The model’s robustness to common EEG noise has to be improved, (2) The temporal, spectral and spatial features of EEG need to be extracted and fused appropriately. Learning both robust and powerful features for MDD diagnosis can improve the overall performance, and multi-task learning is a powerful solution. In this paper, we propose M-MDD, a multi-task deep learning framework for MDD diagnosis using EEG. First, we design the Contrastive Noise Robustness Task to learn noise-independent features. Then, we design the Supervised Feature Extraction Task to extract temporal, spectral and spatial features of EEG respectively, and then effectively combine them together. Finally, the above two modules share the same feature space and are trained jointly with the Multi-task Learning Module, improving the overall performance. Validated on two public MDD diagnosis datasets with subject-independent cross-validation, our model achieves the state-of-the-art performance.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 130008"},"PeriodicalIF":5.5,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143706077","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}
NeurocomputingPub Date : 2025-03-20DOI: 10.1016/j.neucom.2025.130007
Xiaotian Wang , Xiang Jiang , Zhifu Zhao , Kexin Wang , Yifan Yang
{"title":"Exploring interaction: Inner-outer spatial–temporal transformer for skeleton-based mutual action recognition","authors":"Xiaotian Wang , Xiang Jiang , Zhifu Zhao , Kexin Wang , Yifan Yang","doi":"10.1016/j.neucom.2025.130007","DOIUrl":"10.1016/j.neucom.2025.130007","url":null,"abstract":"<div><div>Transformer-based methods have achieved significant results in the field of skeleton-based action recognition. However, when dealing with two-person interaction, existing approaches normally embed the skeleton of each person separately and then introduce an additional module to learn their interactions. This risks losing the spatial and semantic connection information between the two entities, which is crucial for interaction identification. To address this issue, a unified interactive spatial–temporal transformer is proposed in this paper. First, a Two-Person Embedding (TPE) is performed to provide a holistic interactive relationship representation, which can effectively avoid the information gap caused by the division of interacting entities. Second, an innovative Inner-Outer Transformer (IOformer) combining with a new spatio-temporal partition strategy is proposed to simultaneously learn the interactions between intra-partition joints and inter-partition skeletal parts. By comprehensively capturing the key spatio-temporal interactive feature, the accuracy and robustness of interaction recognition can be significantly improved. Extensive experiments on three challenging benchmark datasets validate that our method achieves better performance in comprehensive evaluation methods.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 130007"},"PeriodicalIF":5.5,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143706020","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}
NeurocomputingPub Date : 2025-03-20DOI: 10.1016/j.neucom.2025.130009
Ruiqi Zha, Zhichao Lian, Qianmu Li
{"title":"Centroid-based Contrastive Consistency Learning for transferable deepfake detection","authors":"Ruiqi Zha, Zhichao Lian, Qianmu Li","doi":"10.1016/j.neucom.2025.130009","DOIUrl":"10.1016/j.neucom.2025.130009","url":null,"abstract":"<div><div>Previous research efforts in deepfake detection mainly concentrated on identifying and differentiating artifacts discernible to humans. Those methods have left a bias in learned models, as they tend to concentrate on the disparities between forged and natural regions from the perspective of a single sample while overlooking consistency within categories from the perspective of the entire sample set, which remains crucial across various real-world applications. Therefore, inspired by contrastive learning, we tackle the deepfake detection problem by learning the invariant representations of both categories. Our proposed method, termed Centroid-based Contrastive Consistency Learning (C3L) method, integrates constraints on representations at both the data preprocessing and feature extraction stages. Specifically, during data preprocessing, we consider both temporal relationships within videos and the latent relationships within synthesis data. We introduce a novel Positive Enhancement Module (PEM) designed to characterize natural and forged samples in a facial semantically irrelevant way, thereby guiding a task-oriented positive pair contrasting strategy. In addition, at the feature extraction stage, we introduce the Margin Feature Simulation Module (MFSM), which leverages the centroid of the natural category to simulate marginal features for both categories. Subsequently, we employ the Supervised Contrastive Margin Loss (SCML), utilizing simulated features to emphasize differences at decision boundaries and optimize the learning process. The effectiveness and robustness of the proposed method have been demonstrated through extensive experiments.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"637 ","pages":"Article 130009"},"PeriodicalIF":5.5,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143715513","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":"Fault-tolerant and attack-tolerant cooperative event-triggered sampled-data security control for synchronization of RDNNs with stochastic actuator failures and random deception attacks","authors":"Feng-Liang Zhao , Zi-Peng Wang , Junfei Qiao , Huai-Ning Wu , Tingwen Huang","doi":"10.1016/j.neucom.2025.130021","DOIUrl":"10.1016/j.neucom.2025.130021","url":null,"abstract":"<div><div>In this article, the fault-tolerant and attack-tolerant cooperative event-triggered sampled-data security (FACETSDS) synchronization problem of space-varying reaction–diffusion neural networks (SVRDNNs) under spatially point measurements (SPMs) with stochastic actuator failures and random deception attacks is investigated. First, to save more communication resources and adapt to the variation of system dynamics subject to stochastic actuator failures and random deception attacks, a FACETSDS control scheme is proposed under SPMs. Second, by constructing a Lyapunov functional and utilizing inequality techniques, some synchronization criteria based on spatial linear matrix inequalities (SLMIs) are derived for SVRDNNs. Then, to solve SLMIs, the FETSDS control for synchronization problem of SVRDNNs under SPMs with stochastic actuator failures and random deception attacks is formulated as an linear matrix inequality feasibility problem. Lastly, the designed FACETSDS synchronization strategy is verified by one numerical example.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 130021"},"PeriodicalIF":5.5,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143695955","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}
NeurocomputingPub Date : 2025-03-19DOI: 10.1016/j.neucom.2025.130000
Cong Li, Qingling Wang
{"title":"Distributed approximate aggregative optimization of multiple Euler–Lagrange systems using only sampling measurements","authors":"Cong Li, Qingling Wang","doi":"10.1016/j.neucom.2025.130000","DOIUrl":"10.1016/j.neucom.2025.130000","url":null,"abstract":"<div><div>This article studies the distributed aggregative optimization for multiple Euler–Lagrange systems over directed networks. First, a new class of auxiliary aggregative variables is proposed that only utilize sampling measurements of adjacent outputs. Then, by selecting a smoothing function, we can gradually integrate the sampling information into new variables within the sampling period. Given the proposed variables, a key theorem is derived to transform the approximate aggregative optimization problem into a regulation problem, such that classical control methods can be utilized to regulate the aggregative variables for more complex dynamics. In addition, an adaptive fuzzy distributed control law is constructed based on aggregative variables, deadzone function and fuzzy system to solve the aggregative optimization for fully actuated Lagrangian agents with bounded disturbance. Finally, a numerical experiment is conducted to demonstrate the validity and effectiveness of the theoretical results.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 130000"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687106","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}
NeurocomputingPub Date : 2025-03-19DOI: 10.1016/j.neucom.2025.129967
Rafael Fernandes Pinheiro , Diego Colón , Rui Fonseca-Pinto
{"title":"A memory failure computational model in Alzheimer-like disease via continuous delayed Hopfield network with Lurie control system based healing","authors":"Rafael Fernandes Pinheiro , Diego Colón , Rui Fonseca-Pinto","doi":"10.1016/j.neucom.2025.129967","DOIUrl":"10.1016/j.neucom.2025.129967","url":null,"abstract":"<div><div>Alzheimer’s disease (AD) is a degenerative neurological condition that impacts millions of individuals across the globe and remains without a healing. In the search for new possibilities of treatments for this terrible disease, this work presents the improved Alzheimer-like disease (IALD) model for memory failure and connects it to a new control technique that establishes a cure for the memory lost, either in biological or in artificial neural networks. For the IALD model, continuous Hopfield neural networks (HNN) with time delay are used. From the healing side, a robust control technique is used, which is based on new discoveries in Lurie control systems. In addition, this paper reviews the development of Alzheimer-like disease (ALD) model, as well as, the relationship of HNN with Lurie system. Simulations are executed to validate the model and to show the efficacy of applying a new theorem from Lurie problem. With the results presented, this work proposes a new conceptual paradigm that could potentially be applied in memory failure treatments in AD, as well as in hardware implemented HNN under adversarial attacks or adverse environmental conditions.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 129967"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687112","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":"CITAL: Counterfactual intervention for temporal action localization with point-level annotation","authors":"Yongxiang Hu , Ziying Xia , Zichong Chen , Thupten Tsering , Jian Cheng , Tashi Nyima","doi":"10.1016/j.neucom.2025.130006","DOIUrl":"10.1016/j.neucom.2025.130006","url":null,"abstract":"<div><div>Point-supervised temporal action localization (PTAL) requires only a timestamp annotated on each action instance for training. Most existing PTAL methods use multiple instances learning (MIL) paradigm that localize actions according to contributions of the snippets to the classification results. The gap between classification and localization tasks causes the models to focus more on clues than pure actions. And the models are prone to localize fake actions when there are biased clues between training and test datasets. Inspired by earlier efforts on causal inference, we propose a counterfactual intervention framework for PTAL, CITAL for short. Counterfactual intervention simulates how models respond to counterfactual inputs that contains the same clues without action instances. Intuitively, we can obtain the real response to the pure actions by comparing responses to the inputs before and after counterfactual intervention. Specifically, we propose a background suppression (BS) block to suppresses the background response by guiding the model pay more attention to action instances rather than clues. To fuse the output scores of the various inputs, we propose a fusing by imitation (FI) strategy that further modifies the scores to have a high response to actions and low response to the background segments, generating more accurate proposals. Besides, we propose a counterfactual example generation (CEG) block to generate counterfactual examples with only clues and background contents based on the point labels and snippet-level action scores. Our approach achieves significant mAP gains on THUMOS14, BEOID and GTEA benchmarks comparing to various CAS-based methods without introducing additional parameters.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"637 ","pages":"Article 130006"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143706300","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}
NeurocomputingPub Date : 2025-03-19DOI: 10.1016/j.neucom.2025.129980
Shuang Wang , Qianwen Lu , Boxing Peng , Yihe Nie , Qingchuan Tao
{"title":"DPEC: Dual-Path Error Compensation for low-light image enhancement","authors":"Shuang Wang , Qianwen Lu , Boxing Peng , Yihe Nie , Qingchuan Tao","doi":"10.1016/j.neucom.2025.129980","DOIUrl":"10.1016/j.neucom.2025.129980","url":null,"abstract":"<div><div>For the task of low-light image enhancement, deep learning-based algorithms have demonstrated superiority and effectiveness compared to traditional methods. However, these methods, primarily based on Retinex theory, tend to overlook the noise and color distortions in input images, leading to significant noise amplification and local color distortions in enhanced results. To address these issues, we propose the Dual-Path Error Compensation (DPEC) method, designed to improve image quality under low-light conditions by preserving local texture details while restoring global image brightness without amplifying noise. DPEC incorporates precise pixel-level error estimation to capture subtle differences and an independent denoising mechanism to prevent noise amplification. We introduce the HIS-Retinex loss to guide DPEC’s training, ensuring the brightness distribution of enhanced images closely aligns with real-world conditions. To balance computational speed and resource efficiency while training DPEC for a comprehensive understanding of the global context, we integrated the VMamba architecture into its backbone. Comprehensive quantitative and qualitative experimental results demonstrate that our algorithm significantly outperforms state-of-the-art methods in low-light image enhancement. The code is publicly available online at <span><span>https://github.com/wangshuang233/DPEC</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"637 ","pages":"Article 129980"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143715504","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}
NeurocomputingPub Date : 2025-03-19DOI: 10.1016/j.neucom.2025.129981
Guoxing Wen , Wenxia Sun , Shuaihua Ma
{"title":"Optimized inverse dead-zone formation control using reinforcement learning for the nonlinear single-integrator dynamic multi-agent system","authors":"Guoxing Wen , Wenxia Sun , Shuaihua Ma","doi":"10.1016/j.neucom.2025.129981","DOIUrl":"10.1016/j.neucom.2025.129981","url":null,"abstract":"<div><div>In this article, an optimized inverse dead-zone formation control using identifier–critic–actor reinforcement learning (RL) is studied for the nonlinear single-integral dynamic multi-agent system (MAS). Since MAS formation is often accompanied with a high energy expenditure, it is very necessary and essential to take optimization as a control design principle. In order to smoothly achieve the optimized MAS formation control, a simplified RL is developed by performing the gradient descent method to a simple positive function, which is equivalent to Hamilton–Jacobi–Bellman (HJB) equation. Furthermore, since the MAS formation cooperation is depended on the information exchange among agents, it is very possible to happen the control dead-zone phenomenon, which makes the actuator without control signal. For eliminating the effect of dead-zone, an adaptive inverse dead-zone method is developed and then is combined with RL for this optimized formation control. In comparison to the conventional inverse dead-zone approach, this design has the less adaptive parameters. Finally, the results of theoretical and simulation demonstrate viability of the proposed method.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 129981"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143686891","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}
NeurocomputingPub Date : 2025-03-19DOI: 10.1016/j.neucom.2025.129990
Chengle Zhou , Zhi He , Liwei Zou , Yunfei Li , Antonio Plaza
{"title":"HSACT: A hierarchical semantic-aware CNN-Transformer for remote sensing image spectral super-resolution","authors":"Chengle Zhou , Zhi He , Liwei Zou , Yunfei Li , Antonio Plaza","doi":"10.1016/j.neucom.2025.129990","DOIUrl":"10.1016/j.neucom.2025.129990","url":null,"abstract":"<div><div>Hyperspectral remote sensing technology has demonstrated its spectral diagnosis advantages in numerous remote sensing observation fields. However, hyperspectral imaging is expensive and less portable compared to RGB imaging. To recover the corresponding hyperspectral image (HSI) from a remote sensing RGB image, this paper proposes a new hierarchical semantic-aware convolutional neural network (CNN)-Transformer (HSACT) for remote sensing image spectral super-resolution (SSR). Particularly, this work aims to reconstruct HSIs from RGB images within the same field of view using a lightweight semantic embedding architecture. Our HSACT consists of the following steps. First, an initial spectrum estimation module (from the RGB image to the HSI) is designed to progressively consider spectral estimation between RGB wavelength-inner and wavelength-outer information. Then, an attention-driven semantic-aware CNN-Transformer is developed to reconstruct the spatial and spectral details of HSI. Specifically, a trainable polymorphic superpixel convolution (PSConv) is proposed to capture features efficiently in the above module. Next, we introduce an information-lossless hierarchical network architecture to link the above modules and achieve end-to-end RGB image SSR through weight sharing. Experimental results on several datasets demonstrated that our HSACT outperforms traditional and advanced SSR methods. The codes of this paper are available from <span><span>https://github.com/chengle-zhou/HSACT</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"636 ","pages":"Article 129990"},"PeriodicalIF":5.5,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143686893","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}