Neural NetworksPub Date : 2025-03-19DOI: 10.1016/j.neunet.2025.107389
Ying Hu , Yanping Chen , Yong Xu
{"title":"A shape composition method for named entity recognition","authors":"Ying Hu , Yanping Chen , Yong Xu","doi":"10.1016/j.neunet.2025.107389","DOIUrl":"10.1016/j.neunet.2025.107389","url":null,"abstract":"<div><div>Large language models (LLMs) roughly encode a sentence into a dense representation (a vector), which mixes up the semantic expression of all named entities within a sentence. So the decoding process is easily overwhelmed by sentence-specific information learned during the pre-training process. It results in seriously performance degeneration in recognizing named entities, especially annotated with nested structures. In contrast to LLMs condensing a sentence into a single vector, our model adopts a discriminative language model to map each sentence into a high-order semantic space. In this space, named entities are decomposed into entity body and entity edge. The decomposition is effective to decode complex semantic structures of named entities. In this paper, a shape composition method is proposed for recognizing named entities. This approach leverages a multi-objective learning neural architecture to simultaneously detect entity bodies and classify entity edges. During training, the dual objectives for body and edge learning guide the deep network to encode more task-relevant semantic information. Our method is evaluated on eight widely used public datasets and demonstrated competitive performance. Analytical experiments show that the strategy of let semantic expressions take its course aligns with the entity recognition task. This approach yields finer-grained semantic representations, which enhance not only NER but also other NLP tasks.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107389"},"PeriodicalIF":6.0,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143674855","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-19DOI: 10.1016/j.neunet.2025.107396
Li Li , Jianyi Liu , Hanguang Xiao , Guanqun Zhou , Qiyuan Liu , Zhicheng Zhang
{"title":"Expert guidance and partially-labeled data collaboration for multi-organ segmentation","authors":"Li Li , Jianyi Liu , Hanguang Xiao , Guanqun Zhou , Qiyuan Liu , Zhicheng Zhang","doi":"10.1016/j.neunet.2025.107396","DOIUrl":"10.1016/j.neunet.2025.107396","url":null,"abstract":"<div><div>Abdominal multi-organ segmentation in computed tomography (CT) scans has exhibited successful applications in numerous real clinical scenarios. Nevertheless, prevailing methods for multi-organ segmentation often necessitate either a substantial volume of datasets derived from a single healthcare institution or the centralized storage of patient data obtained from diverse healthcare institutions. This prevailing approach significantly burdens data labeling and collection, thereby exacerbating the associated challenges. Compared to multi organ annotation labels, single organ annotation labels are extremely easy to obtain and have low costs. Therefor, this work establishes an effective collaborative mechanism between multi organ labels and single organ labels, and proposes an expert guided and partially-labeled data collaboration framework for multi organ segmentation, named EGPD-Seg. Firstly, a reward penalty loss function is proposed under the setting of partial labels to make the model more focused on the targets in single organ labels, while suppressing the influence of unlabeled organs on segmentation results. Then, an expert guided module is proposed to enable the model to learn prior knowledge, thereby enabling the model to obtain the ability to segment unlabeled organs on a single organ labeled dataset. The two modules interact with each other and jointly promote the multi organ segmentation performance of the model under label partial settings. This work has been effectively validated on five publicly available abdominal multi organ segmentation datasets, including internal datasets and invisible external datasets. Code: <span><span>https://github.com/LiLiXJTU/EGPDC-Seg</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107396"},"PeriodicalIF":6.0,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-19DOI: 10.1016/j.neunet.2025.107398
Wukun Zheng , Xiao Ke , Wenzhong Guo
{"title":"Zero-shot 3D anomaly detection via online voter mechanism","authors":"Wukun Zheng , Xiao Ke , Wenzhong Guo","doi":"10.1016/j.neunet.2025.107398","DOIUrl":"10.1016/j.neunet.2025.107398","url":null,"abstract":"<div><div>3D anomaly detection aims to solve the problem that image anomaly detection is greatly affected by lighting conditions. As commercial confidentiality and personal privacy become increasingly paramount, access to training samples is often restricted. To address these challenges, we propose a zero-shot 3D anomaly detection method. Unlike previous CLIP-based methods, the proposed method does not require any prompt and is capable of detecting anomalies on the depth modality. Furthermore, we also propose a pre-trained structural rerouting strategy, which modifies the transformer without retraining or fine-tuning for the anomaly detection task. Most importantly, this paper proposes an online voter mechanism that registers voters and performs majority voter scoring in a one-stage, zero-start and growth-oriented manner, enabling direct anomaly detection on unlabeled test sets. Finally, we also propose a confirmatory judge credibility assessment mechanism, which provides an efficient adaptation for possible few-shot conditions. Results on datasets such as MVTec3D-AD demonstrate that the proposed method can achieve superior zero-shot 3D anomaly detection performance, indicating its pioneering contributions within the pertinent domain.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107398"},"PeriodicalIF":6.0,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679906","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-18DOI: 10.1016/j.neunet.2025.107384
Shuangkang Fang , Weixin Xu , Zipeng Feng , Song Yuan , Yufeng Wang , Yi Yang , Wenrui Ding , Shuchang Zhou
{"title":"Arch-Net: Model conversion and quantization for architecture agnostic model deployment","authors":"Shuangkang Fang , Weixin Xu , Zipeng Feng , Song Yuan , Yufeng Wang , Yi Yang , Wenrui Ding , Shuchang Zhou","doi":"10.1016/j.neunet.2025.107384","DOIUrl":"10.1016/j.neunet.2025.107384","url":null,"abstract":"<div><div>The significant computational demands of Deep Neural Networks (DNNs) present a major challenge for their practical application. Recently, many Application-Specific Integrated Circuit (ASIC) chips have incorporated dedicated hardware support for neural network acceleration. However, the lengthy development cycle of ASIC chips means they often lag behind the latest advances in neural architecture research. For instance, Layer Normalization is not well-supported on many popular chips, and the efficiency of 7 × 7 convolution is significantly lower than the equivalent three 3 × 3 convolution. Therefore, in this paper, we introduce Arch-Net, a neural network framework comprised exclusively of a select few common operators, namely 3 × 3 Convolution, 2 × 2 Max-pooling, Batch Normalization, Fully Connected layers, and Concatenation, which are efficiently supported across the majority of ASIC architectures. To facilitate the conversion of disparate network architectures into Arch-Net, we propose the Arch-Distillation methodology, which incorporates strategies such as Residual Feature Adaptation and Teacher Attention Mechanism. These mechanisms enable effective conversion between different network structures alongside efficient model quantization. The resultant Arch-Net eliminates unconventional network constructs while maintaining robust performance even under sub-8-bit quantization, thereby enhancing compatibility and deployment efficiency. Empirical results from image classification and machine translation tasks demonstrate that using only a few types of operators in Arch-Net can achieve results comparable to those obtained with complex architectures. This provides a new insight for deploying structure-agnostic neural networks on various ASIC chips.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107384"},"PeriodicalIF":6.0,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679921","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-18DOI: 10.1016/j.neunet.2025.107386
Tianyi Liu , Zhaorui Tan , Haochuan Jiang , Kaizhu Huang
{"title":"Stagger Network: Rethinking information loss in medical image segmentation with various-sized targets","authors":"Tianyi Liu , Zhaorui Tan , Haochuan Jiang , Kaizhu Huang","doi":"10.1016/j.neunet.2025.107386","DOIUrl":"10.1016/j.neunet.2025.107386","url":null,"abstract":"<div><div>Medical image segmentation presents the challenge of segmenting various-size targets, demanding the model to effectively capture both local and global information. Despite recent efforts using CNNs and ViTs to predict annotations of different scales, these approaches often struggle to effectively balance the detection of targets across varying sizes. Simply utilizing local information from CNNs and global relationships from ViTs without considering potential significant divergence in latent feature distributions may result in substantial information loss. To address this issue, in this paper, we will introduce a novel Stagger Network (SNet) and argues that a well-designed fusion structure can mitigate the divergence in latent feature distributions between CNNs and ViTs, thereby reducing information loss. Specifically, to emphasize both global dependencies and local focus, we design a Parallel Module to bridge the semantic gap. Meanwhile, we propose the Stagger Module, trying to fuse the selected features that are more semantically similar. An Information Recovery Module is further adopted to recover complementary information back to the network. As a key contribution, we theoretically analyze that the proposed parallel and stagger strategies would lead to less information loss, thus certifying the SNet’s rationale. Experimental results clearly proved that the proposed SNet excels comparisons with recent SOTAs in segmenting on the Synapse dataset where targets are in various sizes. Besides, it also demonstrates superiority on the ACDC and the MoNuSeg datasets where targets are with more consistent dimensions.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"188 ","pages":"Article 107386"},"PeriodicalIF":6.0,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143697956","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-18DOI: 10.1016/j.neunet.2025.107385
Juntao Han , Gang Liu , Zhang Zhang
{"title":"Memristive circuit of emotion with negative feedback based on three primary color model","authors":"Juntao Han , Gang Liu , Zhang Zhang","doi":"10.1016/j.neunet.2025.107385","DOIUrl":"10.1016/j.neunet.2025.107385","url":null,"abstract":"<div><div>Many memristive circuits tend to oversimplify the process of emotion generation as a linear event, disregarding crucial factors such as negative feedback and other regulatory mechanisms. In this paper, a memristive circuit of emotion with negative feedback based on three primary color model is proposed to solve the above problems. The designed circuit is composed of perception modules, synapse modules, central nervous system modules and overt behavior module. It realizes emotion generation, emotion evolution and long-term memory functions based on the neural network circuit with behavioral homeostatic negative feedback function. Meanwhile, the three primary color model of basic emotions is discussed and realized. Any two basic emotions can be mixed to produce a higher order emotion. The memristive circuit, based on the three primary color model as a theoretical foundation, offers valuable insights for the further advancement of neural networks.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107385"},"PeriodicalIF":6.0,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143726158","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-17DOI: 10.1016/j.neunet.2025.107390
Qin Xu , Shan Song , Qihang Wu , Bo Jiang , Bin Luo , Jinhui Tang
{"title":"Multi-level semantic-aware transformer for image captioning","authors":"Qin Xu , Shan Song , Qihang Wu , Bo Jiang , Bin Luo , Jinhui Tang","doi":"10.1016/j.neunet.2025.107390","DOIUrl":"10.1016/j.neunet.2025.107390","url":null,"abstract":"<div><div>Effective visual representation is crucial for image captioning task. Among the existing methods, the grid-based visual encoding methods take fragmented features extracted from the entire image as input, lacking the fine-grained semantic information focused on salient objects. To address this issue, we propose an effective method, namely Multi-Level Semantic-Aware Transformer (MLSAT) for image captioning, to simultaneously focus on contextual details and high-level semantic information centered on salient objects. First, to model the spatial correlations of grids and the semantic interactions of salient objects, we propose the Visual Content Guided Attention (VCGA), which adaptively embeds the relative position relationships of the grids into the visual features based on their visual content and is used as the attention layer of the encoder. Then, in order to enhance the visual representation, we propose the Multi-Level Semantic-Aware (MLSA) module which further models the fine-grained semantic information centered on salient objects. In this module, the primary semantic information is first extracted from the encoder by using the Semantic Information Extractor (SIE), then refined by the Semantic Refiner (SR) and adaptively integrated into the visual representation by the Visual-Semantic Fusion Block (V-SFB). Our MLSAT is extensively evaluated on the MS-COCO dataset and outperforms the state-of-the-art models, with 135.1% CIDEr (c40) on the official online testing server. The source code is available at <span><span>https://github.com/XvZhao147/MLSAT</span><svg><path></path></svg></span></div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107390"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143644933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-17DOI: 10.1016/j.neunet.2025.107375
Xin Li , Zhihong Xia , Hongkun Zhang
{"title":"Cauchy activation function and XNet","authors":"Xin Li , Zhihong Xia , Hongkun Zhang","doi":"10.1016/j.neunet.2025.107375","DOIUrl":"10.1016/j.neunet.2025.107375","url":null,"abstract":"<div><div>We have developed a novel activation function, named the <em>Cauchy Activation Function</em>. This function is derived from the <em>Cauchy Integral Theorem</em> in complex analysis and is specifically tailored for problems requiring high precision. This innovation has led to the creation of a new class of neural networks, which we call (Comple)XNet, or simply XNet.</div><div>We will demonstrate that XNet is particularly effective for high-dimensional challenges such as image classification and solving Partial Differential Equations (PDEs). Our evaluations show that XNet significantly outperforms established benchmarks like MNIST and CIFAR-10 in computer vision, and offers substantial advantages over Physics-Informed Neural Networks (PINNs) in both low-dimensional and high-dimensional PDE scenarios.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"188 ","pages":"Article 107375"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143714779","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-17DOI: 10.1016/j.neunet.2025.107377
Xueqin Chen , Xiaoyu Huang , Qiang Gao , Li Huang , Guisong Liu
{"title":"Enhancing text-centric fake news detection via external knowledge distillation from LLMs","authors":"Xueqin Chen , Xiaoyu Huang , Qiang Gao , Li Huang , Guisong Liu","doi":"10.1016/j.neunet.2025.107377","DOIUrl":"10.1016/j.neunet.2025.107377","url":null,"abstract":"<div><div>Fake news poses a significant threat to society, making the automatic and accurate detection of fake news an urgent task. Various detection cues have been explored in extensive research, with news text content shown to be indispensable as it directly reflects the creator’s intent. Existing paradigms for developing text-centric methods, i.e., small language model (SLM)-based, external knowledge-enhanced, and large language model (LLM)-based approaches, have achieved remarkable improvements. However, each of these paradigms still faces the following challenges: (1) the low generalization ability of SLM-based methods, due to their training on limited and specific knowledge; (2) the extensive retrieval operations required by external knowledge-enhanced methods, both during training and at the inference stage, leading to increased computational costs; and (3) LLMs are prone to hallucinations and less suited for factual reasoning. To address these challenges, we propose LEKD, which combines the strengths of SLMs, external knowledge, and LLMs to enhance text-centric fake news detection. Specifically, LEKD leverages the LLM to generate external knowledge as supplementary information for the training set only and introduces a graph-based semantic-aware feature alignment module to resolve knowledge contradictions, as well as an information bottleneck-based knowledge distillation module to ensure the implicit generation of these features during inference. Extensive experiments conducted on two datasets demonstrate the advantages of LEKD over the baselines.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107377"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143644932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural NetworksPub Date : 2025-03-17DOI: 10.1016/j.neunet.2025.107383
Congyan Lv , Guangliang Liu , Yingnan Pan , Zhijian Hu , Yan Lei
{"title":"Event-based distributed cooperative neural learning control for nonlinear multiagent systems with time-varying output constraints","authors":"Congyan Lv , Guangliang Liu , Yingnan Pan , Zhijian Hu , Yan Lei","doi":"10.1016/j.neunet.2025.107383","DOIUrl":"10.1016/j.neunet.2025.107383","url":null,"abstract":"<div><div>In practical engineering, many systems are required to operate under different constraint conditions due to considerations of system security. Violating these constraints conditions during operation may lead to performance degradation. Additionally, communication among agents is highly dependent on the network, which inevitably imposes a network burden on the control systems. To address these issues, this paper investigates the switching event-triggered distributed cooperative learning control issue for nonlinear multiagent systems with time-vary output constraints. An improved output-dependent universal barrier function with adjustable constraint boundaries is proposed, which can uniformly handle symmetric or asymmetric output constraints without changing the controller structure. Meanwhile, an improved switching event-triggered condition is designed based on neural networks (NNs) weight, which can allow the system to adaptively adjust the NNs weight update frequency according to the performance of the system, thereby saving communication resources. Furthermore, the Padé approximation technique is employed to address the input delay issue and simplify the controller design process. Using Lyapunov stability theory, it is proved that the outputs of all followers converge to a neighborhood around the leader output without violating output constraints, and all signals in the closed-loop system remain ultimately bounded. At last, the availability of the presented approach can be verified through some simulation results.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107383"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143674856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}