检索结果-内蒙古大学图书馆

Enhancing Graph Structure Learning via Motif-Driven Hypergraph Construction

IEEE Transactions on Network science and Engineering 2025年第3期12卷 2333-2344页

作者： Zhao, Jia-Le Zhang, Xian-Jie Ding, Xiao Zhang, Xingyi Zhang, Hai-Feng Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Mathematical Science Hefei230601 China Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Computer Science and Technology Hefei230601 China

Graph neural networks (GNNs), as a cutting-edge technology in deep learning, perform particularly well in various tasks that process graph structure data. However, their foundation on pairwise graphs often limits their capacity to capture latent higher-order topological semantic information. Thus, it is crucial to find a way to extract the latent higher-order information of graphs without missing the lower-order information of the original graph. To address this issue, we here develop a method to construct hypergraph based on motifs, and then a novel neural network framework, named MD-HGNN, is proposed for enhanced graph learning. Specifically, we first utilize motifs of the original graph to construct the hypergraph and eliminate nested structures within the hypergraph to prevent information redundancy. Subsequently, GNNs and hypergraph neural networks (HGNNs) are employed separately to extract the lower-order and higher-order topological semantic information of the graph. Finally, the lower-order and higher-order information are integrated to obtain an embedded representation of graph. Extensive experimental results demonstrate that MD-HGNN preserves the original lower-order graph structure information while effectively extracting higher-order features. Moreover, its performance and robustness are validated across different downstream tasks. © 2013 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Recognition of Micromotion Space Targets at Low SNR Based on Complex-Valued Time Convolutional Attention Denoising Recognition Network

IEEE Transactions on Radar Systems

引用

IEEE Transactions on radar Systems 2025年 3卷 193-202页

作者： Bai, Xueru Mao, Xuchen Tian, Xudong Zhou, Feng Xidian University National Key Laboratory of Radar Signal Processing Xi'an710071 China Beijing Institute of Astronautical Systems Engineering Beijing100076 China Xidian University School of Aerospace Science and Technology Xi'an710071 China

For a micromotion space target, its narrowband radar cross section (RCS) series reflects the characteristics of target shape and motion. In practical scenarios, however, the RCS series of distant targets with weak scattering coefficients suffers from low signal-to-noise ratio (SNR), and performing separate noise suppression and recognition purely on the amplitude results in degraded recognition performance. To tackle this issue, an end-to-end complex-valued (CV) time convolutional attention denoising recognition network, dubbed as CV-TCANet, is proposed. Specifically, the denoising module captures temporal correlation by the CV attention mechanism and calculates the noise mask for denoising;and the recognition module utilizes the CV temporal convolutional network (CV-TCN) for feature extraction and recognition. In addition, a hybrid loss is designed to realize the integration of denoising and recognition, thus preserving target information while denoising and improving the recognition accuracy. Experimental results have proved that the proposed method could achieve satisfying recognition performance at low SNR. © 2025 IEEE.

关键词： signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

TBIA-DBNet: A Two-Branch Image-Adaptive DBNet for Scene Text Detection in Real-World Foggy Scenes 27th

TBIA-DBNet: A Two-Branch Image-Adaptive DBNet for Scene Tex...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Liu, Zhaoxi Zhou, Gang He, Runlin Zhang, Mengnan Jia, Zhenhong Ma, Jing Key Laboratory of Signal Detection and Processing Department of Computer Science and Technology Xinjiang University Urumqi China

ISBN: (纸本)9783031781186

Though deep learning-based scene text detection methods have achieved promising results on conventional datasets, these methods are unable to maintain optimal performance in adverse weather conditions, such as foggy weather. To alleviate this problem, we propose a Two-Branch Image-Adaptive DBNet (TBIA-DBNet) framework. Specifically, to avoid missing discriminable features from the original image in one branch, we design an Image Enhancement Network (IENet) in another branch. Additionally, we design a Fusion Module based on Coordinate Attention (FMCA) to fully integrate original and enhanced features. Experimental results demonstrate that TBIA-DBNet significantly enhances scene text detection performance in foggy weather. Notably, it improves detection accuracy by nearly 10% in real-world foggy weather conditions compared to existing methods. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

E-RNS: Enhancing Negative Sample Quality from Gradient Perspective for Graph Recommendation

E-RNS: Enhancing Negative Sample Quality from Gradient Persp...

引用

2025 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Feng, Qiangsheng Qin, Jiwei Ma, Jie School of Computer Science and Technology Xinjiang University China Key Laboratory of Signal Detection and Processing China

ISBN: (纸本)9798350368741

Bayesian Personalized Ranking (BPR) is a widely used optimization function in GNN-based recommender systems, and negative samples are usually obtained through the Random Negative Sampling (RNS) method during BPR training. However, from the gradient perspective, RNS tends to select low-quality samples with minimal information, resulting in small gradients. These small gradients contribute little to BPR optimization, limiting the model's ability to effectively distinguish between positive and negative samples. To alleviate this issue, we propose a general negative sample information enhancement method: Enhancing Random Negative Sampling (E-RNS), which constructs hard negative samples by enhancing the information in randomly selected negative samples. Specifically, in the Noise Injection step, it generates initial noise and injects a certain amount of noise in the same direction into the vector dimensions of positive samples to create enriched information. Then, in the Information Fusion step, this enriched information is mixed with the negative samples to synthesize new hard negative samples. Extensive experiments demonstrate that applying E-RNS to GNN-based recommender models significantly improves performance. © 2025 IEEE.

关键词： Collaborative Filtering Graph Neural Networks Negative Sampling Recommender Systems

来源：评论

学校读者我要写书评

暂无评论

Hyperparameter free sparse estimation for MIMO radar direction finding in the presence of unknown gain-phase errors 16

Hyperparameter free sparse estimation for MIMO radar directi...

引用

16th International Conference on signal processing Systems, ICSPS 2024

作者： Xiao, Jiong Huang, Ping Tang, Bo Zuo, Lei College of Electronic Engineering National University of Defense Technology Hefei China National Key Laboratory of Radar Signal Processing Xidian University Xi’an China

ISBN: (纸本)9781510689251

To improve the accuracy of direction of arrival (DOA) estimation in multiple-input-multiple-output (MIMO) radar systems under unknown gain-phase errors, we propose a gain-phase error calibration and DOA estimation algorithm based on sparse learning via iterative minimization (SLIM). The proposed algorithm harnesses the spatial sparsity of the target signals, and estimates the spatial spectrum and gain-phase error matrices of MIMO arrays through a cyclic optimization procedure. Numerical results show that even in the presence of unknown gain-phase uncertainties, the proposed algorithm can accurately estimate the target DOA using a limited number of samples. © 2025 SPIE.

关键词： Direction of arrival

来源：评论

学校读者我要写书评

暂无评论

Deformable Multi-Scale Network for Snow Removal in Video 27th

Deformable Multi-Scale Network for Snow Removal in Video

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： He, Runlin Zhou, Gang Xue, Tianhao Liu, Zhaoxi Jia, Zhenhong Key Laboratory of Signal Detection and Processing Department of Computer Science and Technology Xinjiang University Urumqi China

ISBN: (纸本)9783031781247

Snowfall severely degrades outdoor video visibility while reducing the performance of subsequent vision tasks. Although video recovery methods based on deep learning have achieved amazing accomplishments, video snow removal still faces problems such as varying scales and intricate trajectories of snowflakes, which makes it difficult to remove snowflakes and easy to create artifacts on moving objects. To address these issues, we propose a deformable multi-scale video desnowing network. Specifically, we design a multi-scale pseudo-3D residual block(MSRB-P3D) that can effectively remove snowflakes of different scales. Furthermore, a deformable large kernel attention 3Dblock(D-LKA 3Dblock) is introduced to capture the inter-frame dynamic information and reduce the artifacts. Due to the scarcity of dataset, we proposed a new dataset named Synthetic and Real Snowy Video Dataset(SRSVD). Extensive experiments have proven that our proposed method not only outperforms other state-of-the-art methods on both synthetic and real snowy videos, but also effectively improves performance on subsequent vision task. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Snow

来源：评论

学校读者我要写书评

暂无评论

WDE-TransUNet: Brain Tumor Segmentation in MRI Images Utilizing Detail Enhancement and Transformers 24

WDE-TransUNet: Brain Tumor Segmentation in MRI Images Utiliz...

引用

5th International Symposium on Artificial Intelligence for Medicine science, ISAIMS 2024

作者： Ni, Xinhu Mo, Jiaqing Xinjiang Key Laboratory of Signal Detection and Processing School of Computer Science and Technology Xinjiang Urumqi China

ISBN: (纸本)9798400717826

WDE-TransUNet is a novel neural network architecture specifically designed for robust and precise semantic segmentation in medical imaging. This method addresses the challenges posed by inconsistent image quality, diverse pathologies, and anatomical variations due to different imaging techniques by integrating Transformer and convolutional neural network (CNN) modules within an encoder-decoder framework. It utilizes prior knowledge to assign weights at the input stage, optimizing feature utilization across modalities and mitigating the impact of irrelevant features. The encoder employs detail enhanced convolution (DEConv) and root mean square (RMS)Transformer modules to enhance local feature extraction, and then conveys these features to the decoder via skip connections, enriching both spatial and channel information. Evaluated on the Brain Tumor Segmentation Challenge (BraTS)2019, 2020, and 2021 datasets, this method achieved outstanding performance, with average Dice scores of 84.74%, 86.64%, and 90.66%, respectively. Our experimental results demonstrate that WDE-TransUNet, by integrating RSM Transformer with CNN in an encoder-decoder configuration, is superior to existing algorithms for brain tumor segmentation. © 2024 Copyright held by the owner/author(s).

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

A Multiscale Convolution SAR Image Target Recognition Method Based on Complex-Valued Neural Networks

引用

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2025年 18卷 10657-10673页

作者： Hou, Guangyu Xin, Zhihui Liao, Guisheng Huang, Penghui Huang, Yuhao Zou, Rui Yunnan Normal University Yunnan Key Laboratory of Opto-Electronic Information Technology Kunming650500 China Xidian University National Laboratory of Radar Signal Processing Xi'an710071 China Shanghai Jiao Tong University School of Electronic Information and Electrical Engineering Shanghai200240 China

Recent advances in deep learning have driven significant success in synthetic aperture radar (SAR) automatic target recognition, particularly through convolutional neural network (CNN) based classification algorithms. However, SAR images possess distinctive physical scattering properties, owing to their unique imaging mechanism. Many deep learning algorithms rely solely on amplitude information, ignoring phase information, which may result in the loss of information in the original complex-valued SAR image and suboptimal performance. To tackle these problems, this article introduces a SAR target recognition approach based on complex-valued operations, designated as complex-valued residual mish activation and convolution block attention module (CBAM) net (CRMC-Net). The CRMC-Net effectively utilizes the amplitude and phase information in complex-valued SAR data. Specifically, first, the elements of CNN, including the input and output layers, the convolution layers, the activation functions, and the pooling layers, are extended to the complex-valued domain. Second, in order to further enhance the representation ability of the model, multiscale information of the target is extracted through different convolution kernel sizes. Finally, the network constructs many complex-valued operation blocks to enhance the robustness of the designed network, including the complex-valued residual block, complex-valued Mish activation function, and complex-valued CBAM. The experimental results obtained from the moving and stationary target capture and recognition dataset and OpenSARShip2.0 dataset demonstrate that the proposed network model outperforms the traditional real-valued models. It can further reduce the classification error and enhance performance. © 2008-2012 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Identifying important nodes of hypergraph:An improved PageRank algorithm

引用

Chinese Physics B 2025年第4期34卷 162-171页

作者： Yu-Hao Piao Jun-Yi Wang Ke-Zan Li School of Mathematics and Computing Science Guilin University of Electronic TechnologyGuilin 541004China School of Information and Communication Guilin University of Electronic TechnologyGuilin 541004China Guangxi Wireless Broadband Communication and Signal Processing Key Laboratory Guilin University of Electronic TechnologyGuilin 541004China

Hypergraphs can accurately capture complex higher-order relationships,but it is challenging to identify their important *** this paper,an improved PageRank(ImPageRank)algorithm is designed to identify important nodes in a directed *** algorithm introduces the Jaccard similarity of directed *** comparing the numbers of common neighbors between nodes with the total number of their neighbors,the Jaccard similarity measure takes into account the similarity between nodes that are not directly connected,and can reflect the potential correlation between *** improved susceptible–infected(SI)model in directed hypergraph is proposed,which considers nonlinear propagation mode and more realistic propagation *** addition,some important node evaluation methods are transferred from undirected hypergraphs and applied to directed ***,the ImPageRank algorithm is used to evaluate the performance of the SI model,network robustness and *** of real networks demonstrate the excellent performance of the proposed algorithm and provide a powerful framework for identifying important nodes in directed hypergraphs.

关键词： hypergraph important node PageRank susceptible-infected(SI)model centrality index

来源：评论

学校读者我要写书评

暂无评论

Joint Optimization of Maximum Achievable Rate in SWIPT Systems Assisted by Active STAR-RIS 18th

Joint Optimization of Maximum Achievable Rate in SWIPT Sys...

引用

18th International Conference on Wireless Artificial Intelligent Computing Systems and Applications, WASA 2024

作者： Yang, Junlong Qin, Xizhong Jia, Zhenhong Mao, Lamu School of Computer Science and Technology XJU Xinjiang China Xinjiang Key Laboratory of Signal Detection and Processing Wulumuqi China

ISBN: (纸本)9783031714665

This investigation focuses on a simultaneous wireless information and power transfer (SWIPT) system, significantly enhanced by an active simultaneously transmitting and reflecting reconfigurable intelligent surface (aSTAR-RIS). It extends the degrees of freedom (DoF) for the system and addresses the "double fading" challenge seen in passive RIS frameworks. Our objective is the maximization of the sum achievable rate (SAR), achieved by optimizing the transmit beamforming vector at the base station (BS) and the coefficient and amplification matrices at the aSTAR-RIS. Through an alternating optimization (AO) strategy and fractional programming (FP), a sub-optimal solution is obtained. Simulations demonstrate that our aSTAR-RIS approach results in a 16.5% increase in performance over conventional active RIS configurations and a significant 114% improvement over passive STAR-RIS schemes. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Inductive power transmission

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：