Neurocomputing最新文献_第3页

Cross-modal feature symbiosis for personalized meta-path generation in heterogeneous networks

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

Neurocomputing

Pub Date : 2025-02-28 DOI: 10.1016/j.neucom.2025.129780

Xiaotong Wu, Liqing Qiu, Weidong Zhao

In heterogeneous graph neural networks (HGNNs), the capture of intricate relationships among various types of entities is essential to achieve advanced machine learning applications. Heterogeneous Information Networks (HINs), composed of interconnected multi-type nodes and edges, face significant challenges in managing semantic diversity and inherent heterogeneity. Traditional methods, which rely on manually designed meta-paths, struggle to adapt dynamically to personalized needs and often neglect the integration of structural and attribute features. To address these limitations, this paper introduces the Cross-Modal Symbiotic Meta-Path Generator (CSMPG) framework. CSMPG integrates two key modules: a Cross-Modal State Generation Module that encodes node structure and attribute information into task-aware state vectors and a Personalized Meta-Path Generation Module that dynamically generates and refines meta-paths using reinforcement learning. By leveraging downstream task feedback, CSMPG optimizes path selection to maximize performance. The framework effectively balances cross-modal feature integration and semantic diversity, uncovering impactful meta-paths that are often overlooked by traditional approaches. Experimental results demonstrate that CSMPG consistently enhances recommendation quality and significantly outperforms structure-only and predefined-path-based models.

{"title":"Cross-modal feature symbiosis for personalized meta-path generation in heterogeneous networks","authors":"Xiaotong Wu, Liqing Qiu, Weidong Zhao","doi":"10.1016/j.neucom.2025.129780","DOIUrl":"10.1016/j.neucom.2025.129780","url":null,"abstract":"<div><div>In heterogeneous graph neural networks (HGNNs), the capture of intricate relationships among various types of entities is essential to achieve advanced machine learning applications. Heterogeneous Information Networks (HINs), composed of interconnected multi-type nodes and edges, face significant challenges in managing semantic diversity and inherent heterogeneity. Traditional methods, which rely on manually designed meta-paths, struggle to adapt dynamically to personalized needs and often neglect the integration of structural and attribute features. To address these limitations, this paper introduces the Cross-Modal Symbiotic Meta-Path Generator (CSMPG) framework. CSMPG integrates two key modules: a Cross-Modal State Generation Module that encodes node structure and attribute information into task-aware state vectors and a Personalized Meta-Path Generation Module that dynamically generates and refines meta-paths using reinforcement learning. By leveraging downstream task feedback, CSMPG optimizes path selection to maximize performance. The framework effectively balances cross-modal feature integration and semantic diversity, uncovering impactful meta-paths that are often overlooked by traditional approaches. Experimental results demonstrate that CSMPG consistently enhances recommendation quality and significantly outperforms structure-only and predefined-path-based models.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129780"},"PeriodicalIF":5.5,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143529541","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

Few-shot medical relation extraction via prompt tuning enhanced pre-trained language model

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

Neurocomputing

Pub Date : 2025-02-27 DOI: 10.1016/j.neucom.2025.129752

Guoxiu He , Chen Huang

Medical relation extraction is crucial for developing structured information to support intelligent healthcare systems. However, acquiring large volumes of labeled medical data is challenging due to the specialized nature of medical knowledge and privacy constraints. To address this, we propose a prompt-enhanced few-shot relation extraction (FSRE) model that leverages few-shot and prompt learning techniques to improve performance with minimal data. Our approach introduces a hard prompt concatenated to the original input, enabling contextually enriched learning. We calculate prototype representations by averaging the intermediate states of each relation class in the support set, and classify relations by finding the shortest distance between the query instance and class prototypes. We evaluate our model against existing deep learning based FSRE models using three biomedical datasets: the 2010 i2b2/VA challenge dataset, the CHEMPROT corpus, and the BioRED dataset, focusing on few-shot scenarios with limited training data. Our model demonstrates exceptional performance, achieving the highest accuracy across all datasets in most training configurations under a 3-way-5-shot condition and significantly surpassing the current state-of-the-art. Particularly, it achieves improvements ranging from 1.25% to 11.25% on the 2010 i2b2/VA challenge dataset, 3.4% to 20.2% on the CHEMPROT dataset, and 2.73% to 10.98% on the BioRED dataset compared to existing models. These substantial gains highlight the model’s robust generalization ability, enabling it to effectively handle previously unseen relations during testing. The demonstrated effectiveness of this approach underscores its potential for diverse medical applications, particularly in scenarios where acquiring extensive labeled data is challenging.

医疗关系提取对于开发结构化信息以支持智能医疗系统至关重要。然而，由于医学知识的专业性和隐私限制，获取大量标注医疗数据具有挑战性。为解决这一问题，我们提出了一种提示增强型少量关系提取（FSRE）模型，该模型利用少量关系提取和提示学习技术，以最少的数据提高性能。我们的方法在原始输入的基础上引入了硬提示，从而实现了丰富的上下文学习。我们通过平均支持集中每个关系类的中间状态来计算原型表示，并通过寻找查询实例与类原型之间的最短距离来对关系进行分类。我们使用三个生物医学数据集（2010 i2b2/VA challenge 数据集、CHEMPROT 语料库和 BioRED 数据集）对我们的模型与现有的基于深度学习的 FSRE 模型进行了评估，重点是训练数据有限的少数几个场景。我们的模型表现出了卓越的性能，在 3 路 5 次拍摄条件下的大多数训练配置中，它在所有数据集上都达到了最高的准确率，大大超过了目前最先进的模型。特别是，与现有模型相比，该模型在 2010 i2b2/VA 挑战赛数据集上提高了 1.25% 到 11.25%，在 CHEMPROT 数据集上提高了 3.4% 到 20.2%，在 BioRED 数据集上提高了 2.73% 到 10.98%。这些大幅提升凸显了该模型强大的泛化能力，使其能够在测试过程中有效处理以前未见过的关系。这种方法的有效性凸显了它在各种医疗应用中的潜力，尤其是在获取大量标记数据具有挑战性的情况下。

{"title":"Few-shot medical relation extraction via prompt tuning enhanced pre-trained language model","authors":"Guoxiu He , Chen Huang","doi":"10.1016/j.neucom.2025.129752","DOIUrl":"10.1016/j.neucom.2025.129752","url":null,"abstract":"<div><div>Medical relation extraction is crucial for developing structured information to support intelligent healthcare systems. However, acquiring large volumes of labeled medical data is challenging due to the specialized nature of medical knowledge and privacy constraints. To address this, we propose a prompt-enhanced few-shot relation extraction (FSRE) model that leverages few-shot and prompt learning techniques to improve performance with minimal data. Our approach introduces a hard prompt concatenated to the original input, enabling contextually enriched learning. We calculate prototype representations by averaging the intermediate states of each relation class in the support set, and classify relations by finding the shortest distance between the query instance and class prototypes. We evaluate our model against existing deep learning based FSRE models using three biomedical datasets: the 2010 i2b2/VA challenge dataset, the CHEMPROT corpus, and the BioRED dataset, focusing on few-shot scenarios with limited training data. Our model demonstrates exceptional performance, achieving the highest accuracy across all datasets in most training configurations under a 3-way-5-shot condition and significantly surpassing the current state-of-the-art. Particularly, it achieves improvements ranging from 1.25% to 11.25% on the 2010 i2b2/VA challenge dataset, 3.4% to 20.2% on the CHEMPROT dataset, and 2.73% to 10.98% on the BioRED dataset compared to existing models. These substantial gains highlight the model’s robust generalization ability, enabling it to effectively handle previously unseen relations during testing. The demonstrated effectiveness of this approach underscores its potential for diverse medical applications, particularly in scenarios where acquiring extensive labeled data is challenging.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129752"},"PeriodicalIF":5.5,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534568","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 dynamic-static feature fusion learning network for speech emotion recognition

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

Neurocomputing

Pub Date : 2025-02-27 DOI: 10.1016/j.neucom.2025.129836

Peiyun Xue , Xiang Gao , Jing Bai , Zhenan Dong , Zhiyu Wang , Jiangshuai Xu

Speech is a paramount mode of human communication, and enhancing the quality and fluency of Human-Computer Interaction (HCI) greatly benefits from the significant contribution of Speech Emotion Recognition (SER). Feature representation poses a persistent challenge in SER. A single feature is difficult to adequately represent speech emotion, while directly concatenating multiple features may overlook the complementary nature and introduce interference due to redundant information. Towards these difficulties, this paper proposes a Multi-feature Learning network based on Dynamic-Static feature Fusion (ML-DSF) to obtain an effective hybrid feature representation for SER. Firstly, a Time-Frequency domain Self-Calibration Module (TFSC) is proposed to help the traditional convolutional neural networks in extracting static image features from the Log-Mel spectrograms. Then, a Lightweight Temporal Convolutional Network (L-TCNet) is used to acquire multi-scale dynamic temporal causal knowledge from the Mel Frequency Cepstrum Coefficients (MFCC). At last, both extracted features groups are fed into a connection attention module, optimized by Principal Component Analysis (PCA), facilitating emotion classification by reducing redundant information and enhancing the complementary information between features. For ensuring the independence of feature extraction, this paper adopts the training separation strategy. Evaluating the proposed model on two public datasets yielded a Weighted Accuracy (WA) of 93.33 % and an Unweighted Accuracy (UA) of 93.12 % on the RAVDESS dataset, and 94.95 % WA and 94.56 % UA on the EmoDB dataset. The obtained results outperformed the State-Of-The-Art (SOTA) findings. Meanwhile, the effectiveness of each module is validated by ablation experiments, and the generalization analysis is carried out on the cross-corpus SER tasks.

语音是人类交流的重要方式，而提高人机交互（HCI）的质量和流畅性则极大地得益于语音情感识别（SER）的重要贡献。在 SER 中，特征表示是一个长期的挑战。单一特征难以充分代表语音情感，而直接连接多个特征可能会忽略互补性，并因冗余信息而引入干扰。针对这些难题，本文提出了一种基于动态-静态特征融合的多特征学习网络（ML-DSF），为 SER 获得有效的混合特征表示。首先，本文提出了时频域自校准模块（TFSC），以帮助传统卷积神经网络从 Log-Mel 光谱图中提取静态图像特征。然后，使用轻量级时域卷积网络 (L-TCNet) 从 Mel Frequency Cepstrum Coefficients (MFCC) 中获取多尺度动态时域因果知识。最后，将提取的两组特征输入连接注意模块，通过主成分分析（PCA）进行优化，减少冗余信息，增强特征间的互补信息，从而促进情绪分类。为确保特征提取的独立性，本文采用了训练分离策略。在两个公开数据集上对所提出的模型进行了评估，结果表明在 RAVDESS 数据集上的加权准确率（WA）为 93.33 %，非加权准确率（UA）为 93.12 %；在 EmoDB 数据集上的加权准确率（WA）为 94.95 %，非加权准确率（UA）为 94.56 %。所获得的结果优于最新研究成果（SOTA）。同时，通过消融实验验证了各模块的有效性，并在跨语料库 SER 任务中进行了泛化分析。

{"title":"A dynamic-static feature fusion learning network for speech emotion recognition","authors":"Peiyun Xue , Xiang Gao , Jing Bai , Zhenan Dong , Zhiyu Wang , Jiangshuai Xu","doi":"10.1016/j.neucom.2025.129836","DOIUrl":"10.1016/j.neucom.2025.129836","url":null,"abstract":"<div><div>Speech is a paramount mode of human communication, and enhancing the quality and fluency of Human-Computer Interaction (HCI) greatly benefits from the significant contribution of Speech Emotion Recognition (SER). Feature representation poses a persistent challenge in SER. A single feature is difficult to adequately represent speech emotion, while directly concatenating multiple features may overlook the complementary nature and introduce interference due to redundant information. Towards these difficulties, this paper proposes a Multi-feature Learning network based on Dynamic-Static feature Fusion (ML-DSF) to obtain an effective hybrid feature representation for SER. Firstly, a Time-Frequency domain Self-Calibration Module (TFSC) is proposed to help the traditional convolutional neural networks in extracting static image features from the Log-Mel spectrograms. Then, a Lightweight Temporal Convolutional Network (L-TCNet) is used to acquire multi-scale dynamic temporal causal knowledge from the Mel Frequency Cepstrum Coefficients (MFCC). At last, both extracted features groups are fed into a connection attention module, optimized by Principal Component Analysis (PCA), facilitating emotion classification by reducing redundant information and enhancing the complementary information between features. For ensuring the independence of feature extraction, this paper adopts the training separation strategy. Evaluating the proposed model on two public datasets yielded a Weighted Accuracy (WA) of 93.33 % and an Unweighted Accuracy (UA) of 93.12 % on the <em>RAVDESS</em> dataset, and 94.95 % WA and 94.56 % UA on the <em>EmoDB</em> dataset. The obtained results outperformed the State-Of-The-Art (SOTA) findings. Meanwhile, the effectiveness of each module is validated by ablation experiments, and the generalization analysis is carried out on the cross-corpus SER tasks.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129836"},"PeriodicalIF":5.5,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534570","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

Inverse reinforcement learning by expert imitation for the stochastic linear–quadratic optimal control problem

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

Neurocomputing

Pub Date : 2025-02-27 DOI: 10.1016/j.neucom.2025.129758

Zhongshi Sun , Guangyan Jia

This article studies inverse reinforcement learning (IRL) for the linear–quadratic stochastic optimal control problem, where two agents are considered. A learner agent lacks knowledge of the expert agent’s cost function, but it reconstructs an underlying cost function by observing the expert agent’s states and controls, thereby imitating the expert agent’s optimal feedback control. We initially present a model-based IRL method, which consists of a policy correction and a policy update from the policy iteration in reinforcement learning, as well as a cost function weight reconstruction informed by the inverse optimal control. Afterward, under this scheme, we propose a model-free off-policy IRL method, which requires no system identification, only collecting behavior data from the learner agent and expert agent once during the iteration process. Moreover, the proofs of the method’s convergence, stability, and non-unique solutions are given. Finally, a numerical example and an inverse mean–variance portfolio optimization example are provided to validate the effectiveness of the presented method.

{"title":"Inverse reinforcement learning by expert imitation for the stochastic linear–quadratic optimal control problem","authors":"Zhongshi Sun , Guangyan Jia","doi":"10.1016/j.neucom.2025.129758","DOIUrl":"10.1016/j.neucom.2025.129758","url":null,"abstract":"<div><div>This article studies inverse reinforcement learning (IRL) for the linear–quadratic stochastic optimal control problem, where two agents are considered. A learner agent lacks knowledge of the expert agent’s cost function, but it reconstructs an underlying cost function by observing the expert agent’s states and controls, thereby imitating the expert agent’s optimal feedback control. We initially present a model-based IRL method, which consists of a policy correction and a policy update from the policy iteration in reinforcement learning, as well as a cost function weight reconstruction informed by the inverse optimal control. Afterward, under this scheme, we propose a model-free off-policy IRL method, which requires no system identification, only collecting behavior data from the learner agent and expert agent once during the iteration process. Moreover, the proofs of the method’s convergence, stability, and non-unique solutions are given. Finally, a numerical example and an inverse mean–variance portfolio optimization example are provided to validate the effectiveness of the presented method.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129758"},"PeriodicalIF":5.5,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534571","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

MEAI-Net: Multiview embedding and attention interaction for multivariate time series forecasting

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

Neurocomputing

Pub Date : 2025-02-27 DOI: 10.1016/j.neucom.2025.129769

Chunru Dong , Wenqing Xu , Feng Zhang , Qiang Hua , Yong Zhang

Multivariate time series (MTS) forecasting plays a critical role in diverse societal applications, including stock market analysis and climate change research. While many existing deep learning models have been proven to be effective in MTS forecasting through complex neural network structures and self-attention mechanisms, several challenges remain: (1) insufficient modeling of complex temporal dependencies, (2) limited ability to handle information redundancy and noise, and (3) inadequate capture of periodic characteristics of time series. To address these problems, we propose a Multiview time-dependent feature Embedding and downsampled subsequences Attention Interaction Network (MEAI-Net) for MTS forecasting. First, MEAI-Net adopts a multiview time-dependent feature embedding mechanism to extract various temporal dependency features from the sequences. Second, it reduces redundancy in the temporal sequence features through downsampling. Third, a subsequences cross-attention module is introduced to enhance information exchange between subsequences. Furthermore, we propose the period consistency loss designed to more effectively capture periodic patterns in time series data. Comprehensive experiments conducted on 12 widely used time series datasets demonstrate that MEAI-Net displays promising performance, providing a competitive alternative to current state-of-the-art approaches in MTS forecasting.

{"title":"MEAI-Net: Multiview embedding and attention interaction for multivariate time series forecasting","authors":"Chunru Dong , Wenqing Xu , Feng Zhang , Qiang Hua , Yong Zhang","doi":"10.1016/j.neucom.2025.129769","DOIUrl":"10.1016/j.neucom.2025.129769","url":null,"abstract":"<div><div>Multivariate time series (MTS) forecasting plays a critical role in diverse societal applications, including stock market analysis and climate change research. While many existing deep learning models have been proven to be effective in MTS forecasting through complex neural network structures and self-attention mechanisms, several challenges remain: (1) insufficient modeling of complex temporal dependencies, (2) limited ability to handle information redundancy and noise, and (3) inadequate capture of periodic characteristics of time series. To address these problems, we propose a Multiview time-dependent feature Embedding and downsampled subsequences Attention Interaction Network (MEAI-Net) for MTS forecasting. First, MEAI-Net adopts a multiview time-dependent feature embedding mechanism to extract various temporal dependency features from the sequences. Second, it reduces redundancy in the temporal sequence features through downsampling. Third, a subsequences cross-attention module is introduced to enhance information exchange between subsequences. Furthermore, we propose the period consistency loss designed to more effectively capture periodic patterns in time series data. Comprehensive experiments conducted on 12 widely used time series datasets demonstrate that MEAI-Net displays promising performance, providing a competitive alternative to current state-of-the-art approaches in MTS forecasting.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129769"},"PeriodicalIF":5.5,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534569","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

Self-weighted subspace clustering with adaptive neighbors

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

Neurocomputing

Pub Date : 2025-02-26 DOI: 10.1016/j.neucom.2025.129754

Zhengyan Liu , Huiwen Wang , Lihong Wang , Qing Zhao

Subspace clustering has attracted increasing attention in recent years owing to its ability to process high-dimensional data effectively. However, existing subspace clustering methods often assume that different features are equally important, and on this basis, a similarity matrix is constructed to generate the clustering structure. However, this practice may significantly affect the clustering performance in cases where the importance of different features significantly differs or where many noisy features exist in the original data. To address these challenges, we propose a novel self-weighted subspace clustering method with adaptive neighbors (SWSCAN). A feature weighting scheme is introduced to assign appropriate weights to different features. Then, we use the self-expressive property and adaptive neighbors approach to capture both the global and local structures within the weighted data space. Moreover, we employ the alternating direction method of multipliers (ADMM) to effectively solve the optimization problem of SWSCAN. Empirical results on both synthetic and practical datasets validate that our proposed method outperforms other comparative clustering techniques and can learn appropriate weights for features.

近年来，子空间聚类因其能有效处理高维数据而受到越来越多的关注。然而，现有的子空间聚类方法通常假定不同特征同等重要，并在此基础上构建相似性矩阵来生成聚类结构。然而，在不同特征的重要性存在显著差异或原始数据中存在许多噪声特征的情况下，这种做法可能会严重影响聚类性能。为了应对这些挑战，我们提出了一种新颖的自适应邻域自加权子空间聚类方法（SWSCAN）。我们引入了一种特征加权方案，为不同的特征分配适当的权重。然后，我们利用自表达特性和自适应邻域方法来捕捉加权数据空间内的全局和局部结构。此外，我们还采用了交替方向乘法（ADMM）来有效解决 SWSCAN 的优化问题。在合成数据集和实际数据集上的实证结果验证了我们提出的方法优于其他比较聚类技术，并能为特征学习适当的权重。

{"title":"Self-weighted subspace clustering with adaptive neighbors","authors":"Zhengyan Liu , Huiwen Wang , Lihong Wang , Qing Zhao","doi":"10.1016/j.neucom.2025.129754","DOIUrl":"10.1016/j.neucom.2025.129754","url":null,"abstract":"<div><div>Subspace clustering has attracted increasing attention in recent years owing to its ability to process high-dimensional data effectively. However, existing subspace clustering methods often assume that different features are equally important, and on this basis, a similarity matrix is constructed to generate the clustering structure. However, this practice may significantly affect the clustering performance in cases where the importance of different features significantly differs or where many noisy features exist in the original data. To address these challenges, we propose a novel self-weighted subspace clustering method with adaptive neighbors (SWSCAN). A feature weighting scheme is introduced to assign appropriate weights to different features. Then, we use the self-expressive property and adaptive neighbors approach to capture both the global and local structures within the weighted data space. Moreover, we employ the alternating direction method of multipliers (ADMM) to effectively solve the optimization problem of SWSCAN. Empirical results on both synthetic and practical datasets validate that our proposed method outperforms other comparative clustering techniques and can learn appropriate weights for features.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129754"},"PeriodicalIF":5.5,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534572","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

Passivity for undirected and directed fractional-order complex networks with adaptive output coupling

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

Neurocomputing

Pub Date : 2025-02-26 DOI: 10.1016/j.neucom.2025.129774

Jin-Liang Wang , Chen-Guang Liu , Shun-Yan Ren , Tingwen Huang

In this paper, the passivity for adaptive output coupled fractional-order complex networks (CNs) under undirected and directed topologies is studied by selecting appropriate coupling weight adjustment strategies, based on which the output synchronization for the presented networks is discussed by leveraging the properties of Mittag-Leffler functions and the Laplace transform technique. Utilizing the developed distributed adaptive schemes, several passivity criteria for the fractional-order CNs with output coupling are derived. Moreover, several output synchronization conditions for the output coupled fractional-order CNs are given by employing the acquired passivity results. Ultimately, simulations from numerical examples are utilized to judge the effectiveness for the adaptive laws and the presented criteria.

引用次数: 0

Adaptive multi-model fusion learning for sparse-reward reinforcement learning

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

Neurocomputing

Pub Date : 2025-02-26 DOI: 10.1016/j.neucom.2025.129748

Giseung Park , Whiyoung Jung , Seungyul Han , Sungho Choi , Youngchul Sung

In this paper, we address intrinsic reward generation for sparse-reward reinforcement learning, where the agent receives limited extrinsic feedback from the environment. Traditional approaches to intrinsic reward generation often rely on prediction errors from a single model, where the intrinsic reward is derived from the discrepancy between the model’s predicted outputs and the actual targets. This approach exploits the observation that less-visited state–action pairs typically yield higher prediction errors. We extend this framework by incorporating multiple prediction models and propose an adaptive fusion technique specifically designed for the multi-model setting. We establish and mathematically justify key axiomatic conditions that any viable fusion method must satisfy. Our adaptive fusion approach dynamically learns the best way to combine prediction errors during training, leading to improved learning performance. Numerical experiments validate the effectiveness of our method, showing significant performance gains across various tasks compared to existing approaches.

{"title":"Adaptive multi-model fusion learning for sparse-reward reinforcement learning","authors":"Giseung Park , Whiyoung Jung , Seungyul Han , Sungho Choi , Youngchul Sung","doi":"10.1016/j.neucom.2025.129748","DOIUrl":"10.1016/j.neucom.2025.129748","url":null,"abstract":"<div><div>In this paper, we address intrinsic reward generation for sparse-reward reinforcement learning, where the agent receives limited extrinsic feedback from the environment. Traditional approaches to intrinsic reward generation often rely on prediction errors from a single model, where the intrinsic reward is derived from the discrepancy between the model’s predicted outputs and the actual targets. This approach exploits the observation that less-visited state–action pairs typically yield higher prediction errors. We extend this framework by incorporating multiple prediction models and propose an adaptive fusion technique specifically designed for the multi-model setting. We establish and mathematically justify key axiomatic conditions that any viable fusion method must satisfy. Our adaptive fusion approach dynamically learns the best way to combine prediction errors during training, leading to improved learning performance. Numerical experiments validate the effectiveness of our method, showing significant performance gains across various tasks compared to existing approaches.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129748"},"PeriodicalIF":5.5,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143529604","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

Brain-like contour detector following Retinex theory and Gestalt perception grouping principles

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

Neurocomputing

Pub Date : 2025-02-26 DOI: 10.1016/j.neucom.2025.129765

Rongtai Cai , Helin Que

Numerous contour detection algorithms draw inspiration from biological vision systems. These algorithms imitate the way simple cells extract edges using Gabor filters. They also suppress edges generated by image textures simulating the non-classical receptive fields (NCRFs), thereby popping up the object contours within the image edges. However, these algorithms are not flawless and may yield imperfect results due to noise pollution, unsatisfactory lighting, limitations in image processing algorithms, and likewise. Weak strengths and pixel loss in contour segments are two common issues. In this paper, we provide two strategies to address these challenges. First, we separate the illumination component from the image following Retinex theory, extract the illumination contour using bio-inspired filters, and boost contour strengths by superimposing the illumination contour. Second, we complete object contours by filling small gaps in contours, using a proposed linking likelihood function that is a joint probability of element distance and orientation difference, following Gestalt perceptual grouping principles. Although not performance-oriented, the experimental results show that our endeavors improve the performance of bio-inspired contour detectors. More importantly, we demonstrate the significance of visual computation theories such as the Retinex theory and the Gestalt perception grouping principle for contour detection.

许多轮廓检测算法都是从生物视觉系统中汲取灵感的。这些算法模仿简单细胞使用 Gabor 滤波器提取边缘的方式。它们还能抑制由图像纹理模拟非经典感受野（NCRF）产生的边缘，从而在图像边缘中弹出物体轮廓。然而，这些算法并非完美无瑕，可能会因噪声污染、不理想的光线、图像处理算法的局限性等原因而产生不完美的结果。轮廓段的强度较弱和像素丢失是两个常见问题。在本文中，我们提供了两种策略来应对这些挑战。首先，我们根据 Retinex 理论从图像中分离出光照部分，利用生物启发滤波器提取光照轮廓，并通过叠加光照轮廓来增强轮廓强度。其次，我们根据格式塔感知分组原理，利用所提出的链接似然函数（即元素距离和方向差异的联合概率），通过填补轮廓中的小缝隙来完善物体轮廓。尽管实验结果并不以性能为导向，但它表明我们的努力改善了生物启发轮廓检测器的性能。更重要的是，我们证明了 Retinex 理论和格式塔感知分组原理等视觉计算理论对于轮廓检测的重要意义。

{"title":"Brain-like contour detector following Retinex theory and Gestalt perception grouping principles","authors":"Rongtai Cai , Helin Que","doi":"10.1016/j.neucom.2025.129765","DOIUrl":"10.1016/j.neucom.2025.129765","url":null,"abstract":"<div><div>Numerous contour detection algorithms draw inspiration from biological vision systems. These algorithms imitate the way simple cells extract edges using Gabor filters. They also suppress edges generated by image textures simulating the non-classical receptive fields (NCRFs), thereby popping up the object contours within the image edges. However, these algorithms are not flawless and may yield imperfect results due to noise pollution, unsatisfactory lighting, limitations in image processing algorithms, and likewise. Weak strengths and pixel loss in contour segments are two common issues. In this paper, we provide two strategies to address these challenges. First, we separate the illumination component from the image following Retinex theory, extract the illumination contour using bio-inspired filters, and boost contour strengths by superimposing the illumination contour. Second, we complete object contours by filling small gaps in contours, using a proposed linking likelihood function that is a joint probability of element distance and orientation difference, following Gestalt perceptual grouping principles. Although not performance-oriented, the experimental results show that our endeavors improve the performance of bio-inspired contour detectors. More importantly, we demonstrate the significance of visual computation theories such as the Retinex theory and the Gestalt perception grouping principle for contour detection.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"633 ","pages":"Article 129765"},"PeriodicalIF":5.5,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143526700","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

Person verification and recognition by combining voice signal and online handwritten signature using hyperbolic function based transformer neural network

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

Neurocomputing

Pub Date : 2025-02-26 DOI: 10.1016/j.neucom.2025.129751

Rohitesh Kumar, Rajib Ghosh

With the increasing popularity of smart portable electronic gadgets, various voice based online person verification systems have been developed. However, these systems are susceptible to attacks where an illegitimate individual feeds a recorded voice of a legitimate person, resulting in false confirmations. To overcome this limitation of voice based person verification systems, this article proposes a hyperbolic function based encoded representation of transformer neural network (ERTNN) framework for person verification and recognition by combining online handwritten signature of the genuine person with his/her voice signal. The proposed hyperbolic function based ERTNN framework for person verification and recognition consists of one multi-headed attention layer with positional encoding, seven convolution layers with skip connections, two dense layers, and an output layer. The positional encoding scheme in the proposed hyperbolic function based ERTNN framework has been implemented using hyperbolic

s i n e

and hyperbolic

c o s i n e

functions. The mel-frequency cepstral coefficients (MFCC), MFCC-delta, and MFCC-delta-delta features of the voice signal have been combined with all the temporal points of online handwritten signature sample of a person to make a combined feature matrix. The combined feature matrix of voice signal and online handwritten signature has been fed as an input to the proposed framework to verify and recognize a person corresponding to the input feature matrix. The novelty of this work lies in proposing the hyperbolic function based positional encoding scheme in the ERTNN framework. The experiments have also been carried out using traditional ERTNN framework employing sinusoidal function based positional encoding scheme, learnable positional encoding based ERTNN framework, relative positional encoding based ERTNN framework as well as by removing the positional encoding scheme from the multi-headed attention layer to have a performance comparison with the proposed framework.

{"title":"Person verification and recognition by combining voice signal and online handwritten signature using hyperbolic function based transformer neural network","authors":"Rohitesh Kumar, Rajib Ghosh","doi":"10.1016/j.neucom.2025.129751","DOIUrl":"10.1016/j.neucom.2025.129751","url":null,"abstract":"<div><div>With the increasing popularity of smart portable electronic gadgets, various voice based online person verification systems have been developed. However, these systems are susceptible to attacks where an illegitimate individual feeds a recorded voice of a legitimate person, resulting in false confirmations. To overcome this limitation of voice based person verification systems, this article proposes a hyperbolic function based encoded representation of transformer neural network (ERTNN) framework for person verification and recognition by combining online handwritten signature of the genuine person with his/her voice signal. The proposed hyperbolic function based ERTNN framework for person verification and recognition consists of one multi-headed attention layer with positional encoding, seven convolution layers with skip connections, two dense layers, and an output layer. The positional encoding scheme in the proposed hyperbolic function based ERTNN framework has been implemented using hyperbolic <span><math><mrow><mi>s</mi><mi>i</mi><mi>n</mi><mi>e</mi></mrow></math></span> and hyperbolic <span><math><mrow><mi>c</mi><mi>o</mi><mi>s</mi><mi>i</mi><mi>n</mi><mi>e</mi></mrow></math></span> functions. The mel-frequency cepstral coefficients (MFCC), MFCC-delta, and MFCC-delta-delta features of the voice signal have been combined with all the temporal points of online handwritten signature sample of a person to make a combined feature matrix. The combined feature matrix of voice signal and online handwritten signature has been fed as an input to the proposed framework to verify and recognize a person corresponding to the input feature matrix. <strong>The novelty of this work lies in proposing the hyperbolic function based positional encoding scheme in the ERTNN framework</strong>. The experiments have also been carried out using traditional ERTNN framework employing sinusoidal function based positional encoding scheme, learnable positional encoding based ERTNN framework, relative positional encoding based ERTNN framework as well as by removing the positional encoding scheme from the multi-headed attention layer to have a performance comparison with the proposed framework.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"632 ","pages":"Article 129751"},"PeriodicalIF":5.5,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143511977","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