Neurocomputing最新文献_第10页

Improving multimodal named entity recognition via text-image relevance prediction with large language models 基于大型语言模型的文本-图像关联预测改进多模态命名实体识别

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-10 DOI: 10.1016/j.neucom.2025.130982

Qingyang Zeng , Minghui Yuan , Yueyang Su , Jia Mi , Qianzi Che , Jing Wan

{"title":"Improving multimodal named entity recognition via text-image relevance prediction with large language models","authors":"Qingyang Zeng , Minghui Yuan , Yueyang Su , Jia Mi , Qianzi Che , Jing Wan","doi":"10.1016/j.neucom.2025.130982","DOIUrl":"10.1016/j.neucom.2025.130982","url":null,"abstract":"<div><div>Multimodal Named Entity Recognition (MNER) is a critical task in information extraction, which aims to identify named entities in text-image pairs and classify them into specific types such as person, organization and location. While existing studies have achieved moderate success by fusing visual and textual features through cross-modal attention mechanisms, two major challenges remain: (1) image-text mismatch, where the two modalities are not always semantically aligned in real-world scenarios; and (2) insufficient labeled data, which hampers the model’s ability to learn complex cross-modal associations and limits generalization. To overcome these challenges, we propose a novel framework that leverages the semantic comprehension and reasoning capabilities of Large Language Models (LLMs). Specifically, for the mismatch issue, we employ LLMs to generate the text-image relevance score with inference reason to guide the subsequent modules. Then we design <strong>T</strong>ext-image <strong>R</strong>elationship <strong>P</strong>redicting (TRP) module, which determines the final feature fusion weights based on the relevance score provided by LLMs. To mitigate data scarcity, we prompt LLMs to identify the key entities in text and incorporate them into the original input. Additionally, we design <strong>T</strong>ext-image <strong>R</strong>elevance <strong>F</strong>eatures <strong>L</strong>earning (TRFL) module to construct positive and negative samples based on the relevance score, employing a supervised contrastive learning method to further enhance the model’s ability to extract key features from image-text pairs. Experiments show that our proposed method achieves F1 scores of 75.32 % and 86.65 % on Twitter-2015 and Twitter-2017 datasets, respectively, demonstrating its effectiveness.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"651 ","pages":"Article 130982"},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144634490","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

SEArch: A self-evolving framework for network architecture optimization SEArch：一个自进化的网络架构优化框架

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-10 DOI: 10.1016/j.neucom.2025.130980

Yongqing Liang, Dawei Xiang, Xin Li

{"title":"SEArch: A self-evolving framework for network architecture optimization","authors":"Yongqing Liang, Dawei Xiang, Xin Li","doi":"10.1016/j.neucom.2025.130980","DOIUrl":"10.1016/j.neucom.2025.130980","url":null,"abstract":"<div><div>This paper studies a fundamental network optimization problem that finds a network architecture with optimal performance (low loss) under given resource budgets (small number of parameters and/or fast inference). Unlike existing network optimization approaches such as network pruning, knowledge distillation (KD), and network architecture search (NAS), in this work we introduce a self-evolving pipeline to perform network optimization. In this framework, a simple network iteratively and adaptively modifies its structure by using the guidance from a teacher network, until it reaches the resource budget. An attention module is introduced to transfer the knowledge from the teacher network to the student network. A splitting edge scheme is designed to help the student model find an optimal macro architecture. The proposed framework combines the advantages of pruning, KD, and NAS, and hence, can efficiently generate networks with flexible structure and desirable performance. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet demonstrate that our framework achieves great performance in this network architecture optimization task.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"651 ","pages":"Article 130980"},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144686637","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 data augmentation in named entity recognition 命名实体识别中数据增强的研究

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-10 DOI: 10.1016/j.neucom.2025.130856

Yi Huang , Yuhan Gao , Chengjuan Ren

引用次数: 0

PFRNet: Progressive multi-scale feature fusion and refinement for RGB-D salient object detection PFRNet: RGB-D显著目标检测的渐进式多尺度特征融合与细化

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-10 DOI: 10.1016/j.neucom.2025.130888

Zhengqian Feng , Wei Wang , Mingle Zhou , Wang Li , Yuan Gao , Jiachen Li , Gang Li

{"title":"PFRNet: Progressive multi-scale feature fusion and refinement for RGB-D salient object detection","authors":"Zhengqian Feng , Wei Wang , Mingle Zhou , Wang Li , Yuan Gao , Jiachen Li , Gang Li","doi":"10.1016/j.neucom.2025.130888","DOIUrl":"10.1016/j.neucom.2025.130888","url":null,"abstract":"<div><div>RGB-D salient object detection, through the integration of multi-modal feature information, is adept at generating visually compelling saliency maps. Despite the advancement of various RGB-D salient object detection models, significant challenges such as detection omissions, inaccurate object localization, and false detections persist, particularly in multi-object environments or cluttered backgrounds. To address these issues, we introduce a Progressive Multi-Scale Feature Fusion and Refinement Network (PFRNet) based on an encoder–decoder architecture. During the feature encoding phase, we utilize a dual-stream Pyramid Vision Transformer as the encoder to extract RGB and depth features. Given that low-level features contain detailed spatial information while high-level features encapsulate semantic information, we adopt the Spatial Detail Aggregation Module (SDAM) and the Semantic Feature Enhancement Module (SFEM) to facilitate the cross-modal fusion of these features. In the feature decoding stage, we design a progressive decoder anchored by the Feature Focusing and Refinement Module (FFRM). This decoder incrementally concentrates and refines discriminative information from fused features at multiple scales, simultaneously eliminating redundant content to achieve precise prediction of salient objects. The experimental results show that PFRNet outperforms 14 existing RGB-D salient object detection models across six public datasets, while demonstrating the method’s strong generalization capabilities in RGB-T salient object detection tasks.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"650 ","pages":"Article 130888"},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656746","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

BFP: Balanced filter pruning via knowledge distillation for efficient deployment of CNNs on edge devices BFP：基于知识蒸馏的平衡滤波器剪枝，用于cnn在边缘设备上的有效部署

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-09 DOI: 10.1016/j.neucom.2025.130946

Haikun Zhang , Yajun Liu

{"title":"BFP: Balanced filter pruning via knowledge distillation for efficient deployment of CNNs on edge devices","authors":"Haikun Zhang , Yajun Liu","doi":"10.1016/j.neucom.2025.130946","DOIUrl":"10.1016/j.neucom.2025.130946","url":null,"abstract":"<div><div>Model pruning can reduce the computational cost of convolutional neural networks (CNNs), which enables CNNs to be deployed on edge devices with limited computational resources. However, most existing CNN pruning methods rely on a single global indicator to evaluate the importance of filters, ignoring local feature redundancy, which can easily lead to the loss of key information and affect the performance recovery and generalization of the pruning model. In light of this circumstance, a novel balanced filter pruning (BFP) method is proposed in this paper, connecting global measurement with focused attention. First, the method utilizes the BN layer scaling coefficient to perform global channel evaluation and mines local information redundancy through feature map correlation to achieve dynamic balance between structure compression and information preservation. Next, the weighting of the above two indicators is used as a balanced indicator for assessing the importance of the filters, and pruning is performed according to the set pruning rate. Finally, knowledge distillation is used to compensate for the loss of performance caused by the pruning network, which makes the method show better application prospects in scenarios such as edge computing. The effectiveness of the proposed method is validated on two image classification datasets. For example, for the ResNet-50 on the ImageNet dataset, BFP achieves a 59.2 % reduction in float-point-operations (FLOPs) and a 47.1 % reduction in parameters, and the Top-1 accuracy and Top-5 accuracy of the model only lose 0.52 % and 0.35 %, respectively.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"650 ","pages":"Article 130946"},"PeriodicalIF":5.5,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144655025","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

Federated continual learning: Concepts, challenges, and solutions 联合持续学习：概念、挑战和解决方案

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-09 DOI: 10.1016/j.neucom.2025.130844

Parisa Hamedi , Roozbeh Razavi-Far , Ehsan Hallaji

引用次数: 0

Influence maximization in multilayer social networks using transformer-based node embeddings and deep neural networks 基于变压器节点嵌入和深度神经网络的多层社交网络影响最大化

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-09 DOI: 10.1016/j.neucom.2025.130939

Xilai Ju , Ali Seyfi , Asgarali Bouyer , Alireza Rouhi , Xiaoyang Liu , Bahman Arasteh

{"title":"Influence maximization in multilayer social networks using transformer-based node embeddings and deep neural networks","authors":"Xilai Ju , Ali Seyfi , Asgarali Bouyer , Alireza Rouhi , Xiaoyang Liu , Bahman Arasteh","doi":"10.1016/j.neucom.2025.130939","DOIUrl":"10.1016/j.neucom.2025.130939","url":null,"abstract":"<div><div>Studies have shown that individuals frequently interact and influence other people within a society. Understanding and identifying influential individuals within a network is crucial for optimizing information diffusion. This challenge, known as influence maximization, has attracted significant attention, particularly in multilayer social networks where individuals participate across diverse contexts. Traditional approaches rely on heuristic or approximation algorithms, but their scalability and adaptability remain limited. In this paper, we propose a novel deep neural network architecture to predict and maximize influence in multilayer social networks. The framework utilizes a combination of node- and layer-specific feature embeddings, a transformer encoder for contextual feature integration, and multilayer perceptron (MLPs) for influence regression. The input comprises feature vectors representing nodes and layers, which are encoded and aggregated to approximate the influence of each node within its respective layer. A final aggregation step computes the total influence spread across layers, enabling efficient seed set selection of highly influential nodes. Our method yields favorable outcomes, effectively tackling challenges such as hardware resource requirements, scalability, and runtime performance. Empirical evaluations against state-of-the-art algorithms demonstrate the effectiveness of the proposed model in achieving superior influence spread with reduced computational overhead. This approach proposes new paths for influence maximization in large-scale, multilayer social networks.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"651 ","pages":"Article 130939"},"PeriodicalIF":5.5,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657060","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 few-shot knowledge reasoning method based on three-way partial order structure and prompt learning 基于三向偏序结构和提示学习的少镜头知识推理方法

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-09 DOI: 10.1016/j.neucom.2025.130947

Yuxuan Huang , Enliang Yan , Peiming Zhang , Tianyong Hao

{"title":"A few-shot knowledge reasoning method based on three-way partial order structure and prompt learning","authors":"Yuxuan Huang , Enliang Yan , Peiming Zhang , Tianyong Hao","doi":"10.1016/j.neucom.2025.130947","DOIUrl":"10.1016/j.neucom.2025.130947","url":null,"abstract":"<div><div>As an emerging topic, few-shot knowledge reasoning is of great significance to the advancement of artificial intelligence. KR-POFSA is an approach that employs a three-way partial order structure for knowledge representation, combined with granular computing to achieve few-shot knowledge reasoning. However, it faces a limitation: when using it for knowledge reasoning, if the number of attributes is huge or there are a lot of cross-relationships between the attributes, then it may generate lots of new potential object patterns, most of which are useless. To solve this problem, this paper devises an innovative approach to improve KR-POFSA, which uses LLM to downstream few-shot knowledge reasoning tasks through appropriate prompt design. Specifically, we design a prompt template that guides LLM to output domain knowledge, such as a correlation matrix, and uses thresholds to limit the generation of invalid patterns. Through three experiments—with 8 objects and 9 attributes, 20 objects and 11 attributes, and 23 objects and 12 attributes, respectively—we demonstrate that our method can not only reduce the discovery of invalid attribute granules and object patterns in granular computing by 30 %–50 %, but also may offer practitioners insights into which attributes to prioritize, minimizing empiricism.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"651 ","pages":"Article 130947"},"PeriodicalIF":5.5,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144632466","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

Leveraging subclass learning for improving uncertainty estimation in deep Learning 利用子类学习改进深度学习中的不确定性估计

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-08 DOI: 10.1016/j.neucom.2025.130954

Dimitrios Spanos, Nikolaos Passalis, Anastasios Tefas

{"title":"Leveraging subclass learning for improving uncertainty estimation in deep Learning","authors":"Dimitrios Spanos, Nikolaos Passalis, Anastasios Tefas","doi":"10.1016/j.neucom.2025.130954","DOIUrl":"10.1016/j.neucom.2025.130954","url":null,"abstract":"<div><div>Machine learning is becoming increasingly popular across various applications and has led to state-of-the-art results, but it faces challenges related to its trustworthiness. One aspect of making deep learning models more trustworthy is improving their ability to estimate the uncertainty of whether a sample is from the in-domain (ID) data distribution or not. Especially, neural networks have a tendency to make overly confident extrapolations and struggle to convey their uncertainty, which can limit their trustworthiness. Recent approaches have employed Radial Basis Function (RBF)-based models with great success in improving uncertainty estimation in Deep Learning. However, such models assume a unimodal distribution of the data for each class, which we show is critical for out-of-distribution sample detection, but can be limiting in many real world cases. To overcome these limitations, in this paper, we propose a method for training a deep model utilizing the inherent different modalities that naturally arise in a class in real data, which we call <em>subclasses</em>, leading to improved uncertainty quantification. The proposed method leverages a variance-preserving reconstruction-based representation learning approach that prevents feature collapse and enables robust discovery of subclasses, further improving the effectiveness of the proposed approach. The improvement of the approach is demonstrated using extensive experiments on several datasets.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"651 ","pages":"Article 130954"},"PeriodicalIF":5.5,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144634496","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

Finite-time and fixed-time bipartite synchronization of signed networks with mixed delays 混合延迟签名网络的有限时间和固定时间二部同步

IF 5.5 2区计算机科学

Neurocomputing Pub Date : 2025-07-08 DOI: 10.1016/j.neucom.2025.130942

Yao Tan , Junjian Huang , Wei Zhang , Junren Wang , Shiping Wen , Tingwen Huang

引用次数: 0