Neural Networks最新文献

{"title":"A semantic enhancement-based multimodal network model for extracting information from evidence lists","authors":"Shun Luo,&nbsp;Juan Yu","doi":"10.1016/j.neunet.2025.107387","DOIUrl":"10.1016/j.neunet.2025.107387","url":null,"abstract":"<div><div>Courts require the extraction of crucial information about various cases from heterogeneous evidence lists for knowledge-driven decision-making. However, traditional manual screening is complex and inaccurate when confronted with massive evidence lists and cannot meet the demands of legal judgment. Therefore, we propose a semantic enhancement-based multimodal network model (SEBM) to accurately extract critical information from evidence lists. First, we construct the entity semantic graph based on the differences among entity categories in the text content. Subsequently, we extract the features of multiple modalities within the document by employing distinct methods and guide the fusion of features within each modality to enhance the semantic association among them based on the constructed entity semantic graphs. Furthermore, the improved multimodal self-attention mechanism is employed to enhance the interactions between the various modal features, and the loss function combining Taylor polynomials and supervised contrast learning is utilized to reduce the information loss. Finally, SEBM is evaluated using the authentic Chinese evidence list dataset, which includes extensive entity details from diverse case types across multiple law firms. Results from experiments conducted on the authentic evidence list dataset demonstrate that our model performs better than other high-performing models.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107387"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143726159","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}

{"title":"Spatial and frequency information fusion transformer for image super-resolution","authors":"Yan Zhang,&nbsp;Fujie Xu,&nbsp;Yemei Sun,&nbsp;Jiao Wang","doi":"10.1016/j.neunet.2025.107351","DOIUrl":"10.1016/j.neunet.2025.107351","url":null,"abstract":"<div><div>Previous works have indicated that Transformer-based models bring impressive image reconstruction performance in single image super-resolution (SISR). However, existing Transformer-based approaches utilize self-attention within non-overlapping windows. This restriction hinders the network’s ability to adopt large receptive fields, which are essential for capturing global information and establishing long-distance dependencies, especially in the early layers. To fully leverage global information and activate more pixels during the image reconstruction process, we have developed a Spatial and Frequency Information Fusion Transformer (SFFT) with an expansive receptive field. SFFT concurrently combines spatial and frequency domain information to comprehensively leverage their complementary strengths, capturing both local and global image features while integrating low and high-frequency information. Additionally, we utilize the overlapping cross-attention block (OCAB) to facilitate pixel transmission between adjacent windows, enhancing network performance. During the training stage, we incorporate the Fast Fourier Transform (FFT) loss, thereby fully leveraging the capabilities of our proposed modules and further tapping into the model’s potential. Extensive quantitative and qualitative evaluations on benchmark datasets indicate that the proposed algorithm surpasses state-of-the-art methods in terms of accuracy. Specifically, our method achieves a PSNR score of 32.67 dB on the Manga109 dataset, surpassing SwinIR by 0.64 dB and HAT by 0.19 dB, respectively. The source code and pre-trained models are available at <span><span>https://github.com/Xufujie/SFFT</span><svg><path></path></svg></span></div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107351"},"PeriodicalIF":6.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143644927","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}

{"title":"Deformation-invariant neural network and its applications in distorted image restoration and analysis","authors":"Han Zhang ,&nbsp;Qiguang Chen ,&nbsp;Lok Ming Lui","doi":"10.1016/j.neunet.2025.107378","DOIUrl":"10.1016/j.neunet.2025.107378","url":null,"abstract":"<div><div>Images degraded by geometric distortions pose a significant challenge to imaging and computer vision tasks such as object recognition. Deep learning-based imaging models usually fail to give accurate performance for geometrically distorted images. In this paper, we propose the deformation-invariant neural network (DINN), a framework to address the problem of imaging tasks for geometrically distorted images. The DINN outputs consistent latent features for images that are geometrically distorted but represent the same underlying object or scene. The idea of DINN is to incorporate a simple component, called the quasiconformal transformer network (QCTN), into other existing deep networks for imaging tasks. The QCTN is a deep neural network that outputs a quasiconformal map, which can be used to transform a geometrically distorted image into an improved version that is closer to the distribution of natural or good images. It first outputs a Beltrami coefficient, which measures the quasiconformality of the output deformation map. By controlling the Beltrami coefficient, the local geometric distortion under the quasiconformal mapping can be controlled. The QCTN is lightweight and simple, which can be readily integrated into other existing deep neural networks to enhance their performance. Leveraging our framework, we have developed an image classification network that achieves accurate classification of distorted images. Our proposed framework has been applied to restore geometrically distorted images by atmospheric turbulence and water turbulence. DINN outperforms existing GAN-based restoration methods under these scenarios, demonstrating the effectiveness of the proposed framework. Additionally, we apply our proposed framework to the 1-1 verification of human face images under atmospheric turbulence and achieve satisfactory performance, further demonstrating the efficacy of our approach.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107378"},"PeriodicalIF":6.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679905","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}

{"title":"Improving generalization of neural Vehicle Routing Problem solvers through the lens of model architecture","authors":"Yubin Xiao ,&nbsp;Di Wang ,&nbsp;Xuan Wu ,&nbsp;Yuesong Wu ,&nbsp;Boyang Li ,&nbsp;Wei Du ,&nbsp;Liupu Wang ,&nbsp;You Zhou","doi":"10.1016/j.neunet.2025.107380","DOIUrl":"10.1016/j.neunet.2025.107380","url":null,"abstract":"<div><div>Neural models produce promising results when solving Vehicle Routing Problems (VRPs), but may often fall short in generalization. Recent attempts to enhance model generalization often incur unnecessarily large training cost or cannot be directly applied to other models solving different VRP variants. To address these issues, we take a novel perspective on model architecture in this study. Specifically, we propose a plug-and-play Entropy-based Scaling Factor (ESF) and a Distribution-Specific (DS) decoder to enhance the size and distribution generalization, respectively. ESF adjusts the attention weight pattern of the model towards familiar ones discovered during training when solving VRPs of varying sizes. The DS decoder explicitly models VRPs of multiple training distribution patterns through multiple auxiliary light decoders, expanding the model representation space to encompass a broader range of distributional scenarios. We conduct extensive experiments on both synthetic and widely recognized real-world benchmarking datasets and compare the performance with seven baseline models. The results demonstrate the effectiveness of using ESF and DS decoder to obtain a more generalizable model and showcase their applicability to solve different VRP variants, i.e., traveling salesman problem and capacitated VRP. Notably, our proposed generic components require minimal computational resources, and can be effortlessly integrated into conventional generalization strategies to further elevate model generalization.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107380"},"PeriodicalIF":6.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143644934","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}

{"title":"Bilinear Spatiotemporal Fusion Network: An efficient approach for traffic flow prediction","authors":"Jing Chen ,&nbsp;Shixiang Pan ,&nbsp;Weimin Peng ,&nbsp;Wenqiang Xu","doi":"10.1016/j.neunet.2025.107382","DOIUrl":"10.1016/j.neunet.2025.107382","url":null,"abstract":"<div><div>Accurate traffic flow forecasting is critical for intelligent transportation systems, yet increasing model complexity in spatiotemporal graph neural networks does not always yield proportional gains. In this paper, we present a Bilinear Spatiotemporal Fusion Network (BLSTF) tailored for stable, periodic traffic scenarios. First, a temporal enhancement module is introduced to mitigate multi-step error accumulation. Second, predefined graph priors with linear feedback leverage known road topologies for straightforward yet effective spatial modeling. Finally, a bilinear fusion mechanism seamlessly integrates refined temporal and spatial features with minimal computational overhead. Extensive experiments on four real-world datasets show that BLSTF outperforms state-of-the-art methods, achieving MAE and MAPE of 14.05 and 13.90% on PEMS03, 17.93 and 12.12% on PEMS04, 18.87 and 7.86% on PEMS07, and 13.49 and 8.71% on PEMS08, demonstrating BLSTF’s potential to deliver accurate, efficient, and interpretable traffic flow forecasts.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107382"},"PeriodicalIF":6.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143644154","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}

{"title":"A novel self-supervised graph clustering method with reliable semi-supervision","authors":"Weijia Lu ,&nbsp;Min Wang ,&nbsp;Yun Yu ,&nbsp;Liang Ma ,&nbsp;Yaxiang Shi ,&nbsp;Zhongqiu Huang ,&nbsp;Ming Gong","doi":"10.1016/j.neunet.2025.107418","DOIUrl":"10.1016/j.neunet.2025.107418","url":null,"abstract":"<div><div>Cluster analysis, as a core technique in unsupervised learning, has widespread applications. With the increasing complexity of data, deep clustering, which integrates the advantages of deep learning and traditional clustering algorithms, demonstrates outstanding performance in processing high-dimensional and complex data. However, when applied to graph data, deep clustering faces two major challenges: noise and sparsity. Noise introduces misleading connections, while sparsity makes it difficult to accurately capture relationships between nodes. These two issues not only increase the difficulty of feature extraction but also significantly affect clustering performance. To address these problems, we propose a novel Self-Supervised Graph Clustering model based on Reliable Semi-Supervision (SSGC-RSS). This model innovates through upstream and downstream components. The upstream component employs a dual-decoder graph autoencoder with joint clustering optimization, preserving latent information of features and graph structure, and alleviates the sparsity problem by generating cluster centers and pseudo-labels. The downstream component utilizes a semi-supervised graph attention encoding network based on highly reliable samples and their pseudo-labels to select reliable samples for training, thereby effectively reducing the interference of noise. Experimental results on multiple graph datasets demonstrate that, compared to existing methods, SSGC-RSS achieves significant performance improvements, with accuracy improvements of 0.9%, 2.0%, and 5.6% on Cora, Citeseer, and Pubmed datasets respectively, proving its effectiveness and superiority in complex graph data clustering tasks.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107418"},"PeriodicalIF":6.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679996","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}

{"title":"CFI-Former: Efficient lane detection by multi-granularity perceptual query attention transformer","authors":"Rong Gao ,&nbsp;Siqi Hu ,&nbsp;Lingyu Yan ,&nbsp;Lefei Zhang ,&nbsp;Jia Wu","doi":"10.1016/j.neunet.2025.107347","DOIUrl":"10.1016/j.neunet.2025.107347","url":null,"abstract":"<div><div>Benefiting from the booming development of Transformer methods, the performance of lane detection tasks has been rapidly improved. However, due to the influence of inaccurate lane line shape constraints, the query sequences of existing transformer-based lane line detection methods contain a large number of repetitive and invalid information regions, which leads to redundant information in the detection region and makes the processing of information on localized feature details of the lanes biased. In this paper, a multi-granularity perceptual query attention transformer lane detection method, CFI-Former, is proposed to achieve more accurate lane detection. Specifically, a multi-granularity perceptual query attention (GQA) module is designed to extract lane local detail information. By a two-stage query from coarse to fine, redundant key–value pairs with low information relevance are first filtered out, and then fine-grained token-to-token attention is executed on the remaining candidate regions. This module emphasizes the multi-granularity nuances of lane features from global to local, leading to more effective models based on lane line shape constraints. In addition, weighted adaptive LIoU loss (<span><math><msub><mrow><mi>L</mi></mrow><mrow><mi>φ</mi><mo>−</mo><mi>L</mi><mi>I</mi><mtext>oU</mtext></mrow></msub></math></span>) is proposed to improve lane detection in more challenging scenarios by adaptively increasing the relative gradient of high IoU lane objects and the weight of the loss. Extensive experiments show that CFI-Former outperforms the baseline on two popular lane detection benchmark datasets.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107347"},"PeriodicalIF":6.0,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143636717","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}

{"title":"EBM-WGF: Training energy-based models with Wasserstein gradient flow","authors":"Ben Wan ,&nbsp;Cong Geng ,&nbsp;Tianyi Zheng ,&nbsp;Jia Wang","doi":"10.1016/j.neunet.2025.107300","DOIUrl":"10.1016/j.neunet.2025.107300","url":null,"abstract":"<div><div>Energy-based models (EBMs) show their efficiency in density estimation. However, MCMC sampling in traditional EBMs suffers from expensive computation. Although EBMs with minimax game avoid the above drawback, the energy estimation and generator’s optimization are not always stable. We find that the reason for this instability arises from the inaccuracy of minimizing KL divergence between generative and energy distribution along a vanilla gradient flow. In this paper, we leverage the Wasserstein gradient flow (WGF) of the KL divergence to correct the optimization direction of the generator in the minimax game. Different from existing WGF-based models, we pullback the WGF to parameter space and solve it with a variational scheme for bounded solution error. We propose a new EBM with WGF that overcomes the instability of the minimax game and avoids computational MCMC sampling in traditional methods, as we observe that the solution of WGF in our approach is equivalent to Langevin dynamic in EBMs with MCMC sampling. The empirical experiments on toy and natural datasets validate the effectiveness of our approach.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107300"},"PeriodicalIF":6.0,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143637712","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}

{"title":"LUNETR: Language-Infused UNETR for precise pancreatic tumor segmentation in 3D medical image","authors":"Ziyang Shi ,&nbsp;Ruopeng Zhang ,&nbsp;Xiajun Wei ,&nbsp;Cheng Yu ,&nbsp;Haojie Xie ,&nbsp;Zhen Hu ,&nbsp;Xili Chen ,&nbsp;Yongzhong Zhang ,&nbsp;Bin Xie ,&nbsp;Zhengmao Luo ,&nbsp;Wanxiang Peng ,&nbsp;Xiaochun Xie ,&nbsp;Fang Li ,&nbsp;Xiaoli Long ,&nbsp;Lin Li ,&nbsp;Linan Hu","doi":"10.1016/j.neunet.2025.107414","DOIUrl":"10.1016/j.neunet.2025.107414","url":null,"abstract":"<div><div>The identification of early micro-lesions and adjacent blood vessels in CT scans plays a pivotal role in the clinical diagnosis of pancreatic cancer, considering its aggressive nature and high fatality rate. Despite the widespread application of deep learning methods for this task, several challenges persist: (1) the complex background environment in abdominal CT scans complicates the accurate localization of potential micro-tumors; (2) the subtle contrast between micro-lesions within pancreatic tissue and the surrounding tissues makes it challenging for models to capture these features accurately; and (3) tumors that invade adjacent blood vessels pose significant barriers to surgical procedures. To address these challenges, we propose LUNETR (Language-Infused UNETR), an advanced multimodal encoder model that combines textual and image information for precise medical image segmentation. The integration of an autoencoding language model with cross-attention enabling our model to effectively leverage semantic associations between textual and image data, thereby facilitating precise localization of potential pancreatic micro-tumors. Additionally, we designed a Multi-scale Aggregation Attention (MSAA) module to comprehensively capture both spatial and channel characteristics of global multi-scale image data, enhancing the model's capacity to extract features from micro-lesions embedded within pancreatic tissue. Furthermore, in order to facilitate precise segmentation of pancreatic tumors and nearby blood vessels and address the scarcity of multimodal medical datasets, we collaborated with Zhuzhou Central Hospital to construct a multimodal dataset comprising CT images and corresponding pathology reports from 135 pancreatic cancer patients. Our experimental results surpass current state-of-the-art models, with the incorporation of the semantic encoder improving the average Dice score for pancreatic tumor segmentation by 2.23 %. For the Medical Segmentation Decathlon (MSD) liver and lung cancer datasets, our model achieved an average Dice score improvement of 4.31 % and 3.67 %, respectively, demonstrating the efficacy of the LUNETR.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107414"},"PeriodicalIF":6.0,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143674857","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}

{"title":"A novel deep transfer learning method based on explainable feature extraction and domain reconstruction","authors":"Li Wang,&nbsp;Lucong Zhang,&nbsp;Ling Feng,&nbsp;Tianyu Chen,&nbsp;Hongwu Qin","doi":"10.1016/j.neunet.2025.107401","DOIUrl":"10.1016/j.neunet.2025.107401","url":null,"abstract":"<div><div>Although deep transfer learning has made significant progress, its “black-box” nature and unstable feature adaptation remain key obstacles. This study proposes a multi-stage deep transfer learning method, called XDTL, which combines explainable feature extraction and domain reconstruction to enhance the performance of target models. Specifically, the study first divides features into key and regular features through cross-validation and explainability analysis, then reconstructs the target domain using a seed replacement method based on key target samples, ultimately achieving deep transfer. Experimental results show that, compared to other methods, XDTL achieves an average improvement of 27.43 % in effectiveness, demonstrating superior performance and stronger explainability. This method offers new insights into addressing the explainability challenges in transfer learning and highlights its potential for broader applications across various tasks.</div></div>","PeriodicalId":49763,"journal":{"name":"Neural Networks","volume":"187 ","pages":"Article 107401"},"PeriodicalIF":6.0,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679903","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}