检索结果-内蒙古大学图书馆

17th Asian Conference on computer Vision, ACCV 2024

作者： Xu, Meng Lin, Mingying Ren, Qi Jia, Sen College of Computer Science and Software Engineering Shenzhen University Shenzhen China

ISBN: (纸本)9789819609109

Transformer-based methods have improved the quality of hyperspectral images (HSIs) reconstructed from RGB by effectively capturing their remote relationships. The self-attention mechanisms in existing Transformer models have not fully considered the spatial sparsity and spectral continuity characteristics of HSIs and fail to effectively filter out significant features, resulting in lower-quality reconstruction results. This paper proposes a sparse spectral transformer model for HSI reconstruction (SSTHyper) to address this limitation, adaptively preserving crucial features. The network consists of an encoder-decoder structure capable of learning shallow and deep spatial-spectral priors, primarily composed of sparse spectral self-attention groups. Introducing a sparse spectral self-attention mechanism allows adaptive masking of non-significant details, enhancing reconstruction accuracy. Meanwhile, a lightweight cross-level fusion network is proposed to reduce model parameters and computational costs to enhance spatial-spectral feature extraction. Experimental results on two benchmark datasets demonstrate the outstanding performance of the proposed method. The code will be released at SSTHyper. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

EchoGCN: An Echo Graph Convolutional Network for Skeleton-Based Action Recognition 27th

EchoGCN: An Echo Graph Convolutional Network for Skeleton-Ba...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Qian, Weiwen Huang, Qian Li, Chang Chen, Zhongqi Mao, Yingchi College of Computer Science and Software Engineering Hohai University Nanjing China

ISBN: (纸本)9783031783531

Graph Convolutional Networks (GCNs) have attracted considerable attention in the realm of human action recognition. However, conventional GCNs-based methods typically struggle to construct adjacency matrices that capture diverse semantics, thus leading to performance limitations. To tackle this issue, we propose the Echo Graph as a set of adjacency matrices, which includes both hierarchical and global graphs. Specifically, our hierarchical graph exploits the hierarchical information based on the selected central broadcast nodes, aggregating joints dispersed across considerable physical distances into a unified semantic space. The global graph we construct transcends the limitations imposed by physically defined topological structures, delving into comprehensive information exchange among nodes. Additionally, we propose node activation and hierarchical activation model. These activations aim to prominently highlight crucial nodes and edges for specific samples. Finally, we incorporate Margin ReLU distillation to improve computational efficiency and design a four-stream integration using only the joint and bone data streams from two central broadcast nodes. Based on the aforementioned components, we propose EchoGCN, capable of extracting representative skeleton features. Experimental results on three datasets (NTU-RGB+D 60, NTU-RGB+D 120, and NW-UCLA) demonstrate that our model achieves state-of-the-art performance. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Musculoskeletal system

来源：评论

学校读者我要写书评

暂无评论

Spectral Modality-Aware Interactive Fusion Network for HSI Super-Resolution 17th

Spectral Modality-Aware Interactive Fusion Network for HSI ...

引用

17th Asian Conference on computer Vision, ACCV 2024

作者： Xu, Meng Mao, Jiayou Mo, Ziqian Fu, Xiyou Jia, Sen College of Computer Science and Software Engineering Shenzhen University Shenzhen China

ISBN: (纸本)9789819609109

Due to the limitations of current spectral imaging equipment in acquiring high-resolution hyperspectral images (HR-HSIs), a common approach is to fuse low-resolution hyperspectral images (LR-HSIs) with high-resolution multispectral images (HR-MSIs). However, most existing methods have not fully taken into account the correlation and discrepancy in modality between hyperspectral images (HSIs) and multispectral images (MSIs). To address this limitation, we propose an innovative spectral modality-aware interactive fusion network (SMIF-NET) for comprehensive extraction of spectral information and seamless feature fusion. First, we introduce the spectral modality-aware transformer (SMAT) with a dual-attention mechanism to compute spectral self-similarity and cross-spectral correlation. Second, we apply the interactive spatial-spectral feature fusion (IS2F2) to fuse the acquired high-level spectral and spatial features. This fusion technique combines spatial-wise and channel-wise squeeze and excitation to achieve seamless integration of spatial-spectral information. Finally, the extensive experiments on three datasets demonstrate the superior performance of SMIF-NET in both visual and quantitative assessments compared to eight state-of-the-art (SOTA) fusion-based methods. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

Real-Time Cross-Domain Gesture and User Identification via COTS WiFi

引用

IEEE Transactions on Mobile Computing 2025年第6期24卷 5124-5137页

作者： Cao, Chenhong Ding, Yue Dai, Miaoling Gong, Wei Zhao, Xibin University of Science and Technology of China School of Computer Science and Technology Langfang101127 China Shanghai University School of Computer Engineering and Science Shanghai200444 China Tsinghua University School of Software Beijing100190 China

WiFi-based gesture recognition has emerged as a promising alternative to computer vision, enabling seamless integration and enhanced interaction in human-computer interaction systems. Simultaneously identifying users during gesture recognition is vital for improving security and personalization. However, existing WiFi-based dual-task recognition approaches often rely on handcrafted features, which hinder precision and introduce delays in cross-domain scenarios. To address these challenges, we propose WiDual, a real-time system for cross-domain gesture recognition and user identification using WiFi signals. By integrating spatial and channel attention mechanisms, WiDual adaptively extracts crucial features for dual-task recognition. The system employs Channel State Information (CSI) visualization to convert WiFi signals into images, facilitating efficient feature extraction and minimizing information loss and latency. Furthermore, a collaborative module fuses gesture and user identity features, enhancing recognition performance. Experimental evaluations on a public dataset with six gestures and six users across diverse environments demonstrate WiDual's effectiveness. It achieves 96% accuracy in cross-domain gesture recognition and 91.27% in user identification. Compared to state-of-the-art methods, WiDual improves user identification accuracy by 26%, gesture recognition by 8%, and reduces processing time sixfold, showcasing its potential for real-time applications. © 2002-2012 IEEE.

关键词： Wireless local area networks (WLAN)

来源：评论

学校读者我要写书评

暂无评论

Clustered Reinforcement Learning

引用

Frontiers of computer science 2025年第4期19卷 43-57页

作者： Xiao MA Shen-Yi ZHAO Zhao-Heng YIN Wu-Jun LI National Key Laboratory for Novel Software Technology Department of Computer Science and TechnologyNanjing UniversityNanjing 210023China Department of Electrical Engineering and Computer Sciences University of CaliforniaBerkeleyCA 94720-1770USA

Exploration strategy design is a challenging problem in reinforcement learning(RL),especially when the environment contains a large state space or sparse *** exploration,the agent tries to discover unexplored(novel)areas or high reward(quality)*** existing methods perform exploration by only utilizing the novelty of *** novelty and quality in the neighboring area of the current state have not been well utilized to simultaneously guide the agent’s *** address this problem,this paper proposes a novel RL framework,called clustered reinforcement learning(CRL),for efficient exploration in *** adopts clustering to divide the collected states into several clusters,based on which a bonus reward reflecting both novelty and quality in the neighboring area(cluster)of the current state is given to the *** leverages these bonus rewards to guide the agent to perform efficient ***,CRL can be combined with existing exploration strategies to improve their performance,as the bonus rewards employed by these existing exploration strategies solely capture the novelty of *** on four continuous control tasks and six hard-exploration Atari-2600 games show that our method can outperform other state-of-the-art methods to achieve the best performance.

关键词： deep reinforcement learning exploration count-based method clustering K-means

来源：评论

学校读者我要写书评

暂无评论

GTE: learning code AST representation efficiently and effectively

引用

science China(Information sciences) 2025年第3期68卷 393-394页

作者： Yihao QIN Shangwen WANG Bo LIN Kang YANG Xiaoguang MAO College of Computer Science and Technology National University of Defense Technology Key Laboratory of Software Engineering for Complex Systems National University of Defense Technology

With the development of deep learning in recent years, code representation learning techniques have become the foundation of many software engineering tasks such as program classification [1] and defect detection. Earlier approaches treat the code as token sequences and use CNN, RNN, and the Transformer models to learn code representations.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Layered Temporal Spatial Graph Attention Reinforcement Learning for Multiplex Networked Industrial Chains Energy Management

引用

Tsinghua science and Technology 2025年第2期30卷 528-542页

作者： Yuanshuang Jiang Kai Di Xingyu Wu Zhongjian Hu Fulin Chen Pan Li Yichuan Jiang School of Computer Science and Engineering Southeast UniversityNanjing 211189China School of Software Southeast UniversityNanjing 211189China School of Cyber Science and Engineering Southeast UniversityNanjing 211189

Demand response has recently become an essential means for businesses to reduce production costs in industrial ***,the current industrial chain structure has also become increasingly complex,forming new characteristics of multiplex networked industrial *** in real-time electricity prices in demand response propagate through the coupling and cascading relationships within and among these network layers,resulting in negative impacts on the overall energy management ***,existing demand response methods based on reinforcement learning typically focus only on individual agents without considering the influence of dynamic factors on intra and inter-network *** paper proposes a Layered Temporal Spatial Graph Attention(LTSGA)reinforcement learning algorithm suitable for demand response in multiplex networked industrial chains to address this *** algorithm first uses Long Short-Term Memory(LSTM)to learn the dynamic temporal characteristics of electricity prices for ***,LTSGA incorporates a layered spatial graph attention model to evaluate the impact of dynamic factors on the complex multiplex networked industrial chain *** demonstrate that the proposed LTSGA approach effectively characterizes the influence of dynamic factors on intra-and inter-network relationships within the multiplex industrial chain,enhancing convergence speed and algorithm performance compared with existing state-of-the-art algorithms.

关键词： demand response multiplex networked industrial chains multiagent reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

CausalConf: Datasize-Aware Configuration Auto-Tuning for Recurring Big Data Processing Jobs via Adaptive Causal Structure Learning

引用

IEEE Transactions on Parallel and Distributed Systems 2025年第7期36卷 1354-1371页

作者： Dou, Hui He, Mingjie Zhang, Lei Zhang, Yiwen Zheng, Zibin Anhui University School of Computer Science and Technology Hefei China Sun Yat-sen University School of Software Engineering Guangzhou China

To ensure high-performance processing capabilities across diverse application scenarios, Big Data frameworks such as Spark and Flink usually provide a number of performance-related parameters to configure. Considering the computation scale and the characteristic of repeated executions of typical recurring Big Data processing jobs, how to automatically tune parameters for performance optimization has emerged as a hot research topic in both academic and industry. With the advantages in interpretability and generalization ability, causal inference-based methods recently prove their advancement over conventional search-based and machine learning-based methods. However, the complexity of Big Data frameworks, the time-varying input dataset size of a recurring job and the limitation of a single causal structure learning algorithm together prevent these methods from practical application. Therefore, in this paper, we design and implement CausalConf, a datasize-aware configuration auto-tuning approach for recurring Big Data processing jobs via adaptive causal structure learning. Specifically, the offline training phase is responsible for training multiple datasize-aware causal structure models with different causal structure learning algorithms, while the online tuning phase is responsible for recommending the next promising configuration in an iterative manner via the Multi-Armed Bandit-based optimal intervention set selection as well as the novel datasize-aware causal Bayesian optimization. To evaluate the performance of CausalConf, a series of experiments are conducted on our local Spark cluster with 9 different previously unknown target applications from HiBench. Experimental results show that the performance speed ratio achieved by CausalConf compared to the four recent and representative baselines can respectively reach 1.45×, 1.31×, 1.26× and 1.54× on average and up to 2.53×, 1.55×, 1.57×, 2.18×. Besides, the average total online tuning cost of CausalConf is reduced b

关键词： Spatio-temporal data

来源：评论

学校读者我要写书评

暂无评论

Optimized COVID-19 detection using sparse deep learning models from multimodal imaging data

引用

Multimedia Tools and Applications 2025年第13期84卷 11451-11474页

作者： Moradi, MohammadMahdi Hassanzadeh, Alireza Haghanifar, Arman Ko, Seok Bum Faculty of Computer Science and Software Engineering Concordia University MontrealQC Canada Faculty of Electrical Engineering Shahid Beheshti University Tehran Tehran Iran Department of Electrical & Computer Engineering University of Saskatchewan SaskatoonSK Canada

This paper explores the global spread of the COVID-19 virus since 2019, impacting 219 countries worldwide. Despite the absence of a definitive cure, the utilization of artificial intelligence (AI) methods for disease diagnosis has demonstrated commendable effectiveness in promptly diagnosing patients and curbing infection transmission. The study introduces a deep learning-based model tailored for COVID-19 detection, leveraging three prevalent medical imaging modalities: computed tomography (CT), chest X-ray (CXR), and Ultrasound. Various deep Transfer Learning Convolutional Neural Network-based (CNN) models have undergone assessment for each imaging modality. For each imaging modality, this study has selected the two most accurate models based on evaluation metrics such as accuracy and loss. Additionally, efforts have been made to prune unnecessary weights from these models to obtain more efficient and sparse models. By fusing these pruned models, enhanced performance has been achieved. The models have undergone rigorous training and testing using publicly available real-world medical datasets, focusing on classifying these datasets into three distinct categories: Normal, COVID-19 Pneumonia, and non-COVID-19 Pneumonia. The primary objective is to develop an optimized and swift model through strategies like Transfer Learning, Ensemble Learning, and reducing network complexity, making it easier for storage and transfer. The results of the trained network on test data exhibit promising outcomes. The accuracy of these models on the CT scan, X-ray, and ultrasound datasets stands at 99.4%, 98.9%, and 99.3%, respectively. Moreover, these models’ sizes have been substantially reduced and optimized by 51.93%, 38.00%, and 69.07%, respectively. This study proposes a computer-aided-coronavirus-detection system based on three standard medical imaging techniques. The intention is to assist radiologists in accurately and swiftly diagnosing the disease, especially during the screen

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Network Traffic Anomaly Detection Based on DGBi-SA Model

引用

engineering Letters 2025年第3期33卷 612-619页

作者： Wang, Xin Dai, Hong Huang, Zheng Han, Yue University of Science and Technology Liaoning Liaoning CO Anshan114051 China School of Computer Science and Software Engineering University of Science and Technology Liaoning Liaoning CO Anshan114051 China

Network traffic anomaly detection plays a crucial role in today's network security and performance management. In response to the challenges in current network traffic data processing, such as insufficient structuring and dynamics, the paper proposes a network traffic anomaly detection approach based on DGBi-SA model, which integrates Dynamic Graph Convolutional Networks (DyGCN), Bi-directional Long Short-Term Memory (Bi-LSTM), and the Self-Attention mechanism. To solve the problem of insufficient structuring, the model employs DyGCN to capture network traffic's dynamic graph structural characteristics proficiently. Simultaneously, combining Bi-LSTM and the Self-Attention mechanism enhances the model's ability to capture temporal and spatial dependencies. The main objective of the research in the paper is to improve the accuracy and detection efficiency of network traffic prediction, thereby achieving effective detection and timely response to abnormal behaviors in complex network environments. Through a sequence of experiments conducted on the CICIDS-2017 dataset, the paper verifies that the DGBi-SA model can effectively solve the above problems. The experimental results indicate that the model outperforms existing methods in key performance indicators such as AUC, precision, recall, and F1 score, thereby proving that the DGBi-SA model has important application value and scientific significance in network traffic anomaly detection. © 2025, International Association of Engineers. All rights reserved.

关键词： Long short-term memory

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：