检索结果-内蒙古大学图书馆

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年第7期47卷 5794-5805页

作者： Chen, Man-Sheng Lai, Pei-Yuan Liao, De-Zhang Wang, Chang-Dong Lai, Jian-Huang Sun Yat-sen University School of Computer Science and Engineering Guangzhou510275 China Guangxi Zhuang Autonomous Region Information Center Guangxi Key Laboratory of Digital Infrastructure Nanning530219 China Guangdong Key Laboratory of Big Data Analysis and Processing Guangzhou510006 China Ministry of Education Key Laboratory of Machine Intelligence and Advanced Computing Beijing100816 China South China Technology Commercialization Center Guangzhou510006 China Guangdong University of Technology School of Information Engineering Guangzhou510006 China Guangdong Key Laboratory of Information Security Technology Guangzhou510033 China Ministry of Education Beijing100044 China

Due to the wide existence of unlabeled graph-structured data (e.g., molecular structures), the graph-level clustering has recently attracted increasing attention, whose goal is to divide the input graphs into several disjoint groups. However, the existing methods habitually focus on learning the graphs embeddings with different graph reguralizations, and seldom refer to the obvious differences in data distributions of distinct graph-level datasets. How to characteristically consider multiple graph-level datasets in a general well-designed model without prior knowledge is still challenging. In view of this, we propose a novel Graph Prompt Clustering (GPC) method. Within this model, there are two main modules, i.e., graph model pretraining as well as prompt and finetuning. In the graph model pretraining module, the graph model is pretrained by a selected source graph-level dataset with mutual information maximization and self-supervised clustering regularization. In the prompt and finetuning module, the network parameters of the pretrained graph model are frozen, and a groups of learnable prompt vectors assigned to each graph-level representation are trained for adapting different target graph-level datasets with various data distributions. Experimental results across six benchmark datasets demonstrate the impressive generalization capability and effectiveness of GPC compared with the state-of-the-art methods. © 1979-2012 IEEE.

关键词： Graph embeddings

来源：评论

学校读者我要写书评

暂无评论

Fiber-based quantum secure direct communication without active polarization compensation

引用

Science China(Physics,Mechanics & Astronomy) 2022年第12期65卷 91-96页

作者： Xin Liu Di Luo Guangshen Lin Zihao Chen Chunfeng Huang Shizhuo Li Chengxian Zhang Zhenrong Zhang Kejin Wei Guangxi Key Laboratory for Relativistic Astrophysics School of Physical Science and TechnologyGuangxi UniversityNanning530004China Guangxi Key Laboratory of Multimedia Communications and Network Technology School of ComputerElectronicsand InformationGuangxi UniversityNanning530004China

Quantum secure direct communication(QSDC)that allows people to directly transmit confidential information through insecure channels is an important branch of quantum *** widespread adoption of the QSDC demands the development of simple and stable ***,most of the existent QSDC systems involve a complex self-alignment process at the initial stage and additional hardware to compensate environmental *** this study,we present a fiber-based QSDC system without active polarization *** system comprises a stable transmitter and a novel Sagnac-Mach-Zehnder interferometer for security *** robust system simplifies the self-alignment and is immune to environmental *** robustness of the system was theoretically and experimentally verified,and low bit error rates in a 12 min continuous operation with an active polarization scrambler were *** addition,we performed a proof-of-principle QSDC demonstration,and a secrecy capacity of 3.43 kbps over a 5 km fiber with a detection bit error rate of 0.85%±0.07%and a quantum bit error rate of 0.42%±0.05%were *** results confirm the viability of the proposed QSDC system for practical applications.

关键词： quantum secure direct communication Sagnac-Mach-Zehnder interferometer without active polarization compensation

来源：评论

学校读者我要写书评

暂无评论

Temporal Refinement Graph Convolutional network for Skeleton-Based Action Recognition

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第4期5卷 1586-1598页

作者： Zhuang, Tianming Qin, Zhen Ding, Yi Deng, Fuhu Chen, Leduo Qin, Zhiguang Raymond Choo, Kim-Kwang University of Electronic Science and Technology of China Network and Data Security Key Laboratory of Sichuan Province Chengdu Sichuan610054 China University of Electronic Science and Technology of China School of Information and Software Engineering Chengdu Sichuan610054 China School of Advanced Technology Suzhou Jiangsu215123 China University of Texas at San Antonio Department of Information Systems and Cyber Security San AntonioTX78249 United States

Human skeleton data, which has served in the aspect of human activity recognition, ought to be the most representative biometric characteristics due to its intuitivity and visuality. The state-of-the-art approaches mainly focus on improving modeling spatial correlations within graph topologies. However, the interframes motional representations are also of vital importance, and we argue that they are worth paying attention to and exploring. Therefore, a temporal refinement module with contrastive learning mechanism is proposed, fuzing as a complementary to the conventional spatial graph convolution layer. In addition, in order to further exploiting the interframe variances and generalizing graph convolutional network (GCN) operation to temporal dimension, a temporal-correlation matrix is introduced to effectively capture dynamic dependencies within frame-pairs, enhancing semantic feature representation. Moreover, since GCN is a typical local operator which lacks of capability to fully model the long-term relations along spatial and temporal variation, to move beyond the limitation, a spatial-temporal cascaded aggregation (STCA) module is designed to enlarge the receptive filter scale. The overall recognition framework consists of three above novelties, which is capable of achieving remarkable performance by evaluating on benchmark datasets (i.e., NTU RGB+D 60, NTU RGB+D 120, PKU-MMD, and Kinetics Skeleton 400). Extensive experiments demonstrate the effectiveness of the proposed framework, e.g., performing recognition accuracy rate of 90.9% and 96.8% on NTU RGB+D 60, 87.9% and 88.9% on NTU RGB+D 120. © 2020 IEEE.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Method and system of power load classification and decomposition based on deep learning 4

Method and system of power load classification and decomposi...

引用

4th International Conference on Smart Grid and Energy Engineering, SGEE 2023

作者： Sun, Hao Yuan, Hongtai Geng, Yudong Chen, Jing Zhang, Chuanfu Li, Wen Qilu University of Technology Shandong Academy of Sciences Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center National Supercomputer Center in Jinan Jinan China Shandong Fundamental Research Center for Computer Science Shandong Provincial Key Laboratory of Computer Networks Jinan China

ISBN: (纸本)9798350359619

This paper introduces a system for the classification and decomposition of electrical loads, addressing the shortcomings of current non-intrusive load decomposition methods that are prone to errors. We propose a method that first classifies electrical loads and then performs decomposition, along with the corresponding system workflow. Initially, we collect household electricity consumption data using a proprietary load collection device developed by the company, constructing an electrical load dataset. Subsequently, we train classification and decomposition methods using this dataset, conducting comparative experiments with various methods to select the most effective one. Based on the trained electrical load classification and decomposition models, we develop an electrical load inference program. The key implementation lies in applying the electrical load classification and decomposition inference program to edge devices. Leveraging the cloud server hosting a backend web service for the mobile app, we interact with users through the app, achieving precise decomposition and monitoring of electrical loads. Integrating edge computing and cloud computing, coupled with a mobile app serving as the interaction platform, this paper presents a high-precision, simplified, and highly scalable solution for electrical load analysis and management. In experimental comparisons, we find that using the Gated Recurrent Unit (GRU) method from deep learning and the Sequence-to-Point Neural network (Seq2point) method outperform other methods in classification and decomposition, significantly improving the accuracy of electrical load classification. Therefore, we select the GRU method and Seq2Point method as the core techniques for the electrical load classification and decomposition system. This system not only demonstrates significant advantages in performance but also possesses high accuracy, model simplification, and scalability, providing an efficient and reliable solution for elec

关键词： Edge computing

来源：评论

学校读者我要写书评

暂无评论

A Global Task Scheduling Method Based on network Measurement and Prediction in Computing Power networks 29

A Global Task Scheduling Method Based on Network Measurement...

引用

29th IEEE International Conference on Parallel and Distributed Systems, ICPADS 2023

作者： Zhang, Chuanfu Li, Wen Chen, Jing Ge, Jing Wang, Di Li, Na Sun, Hao Geng, Yudong Qilu University of Technology Shandong Academy of Sciences Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center National Supercomputer Center in Jinan Jinan China Shandong Fundamental Research Center for Computer Science Shandong Provincial Key Laboratory of Computer Networks Jinan China

ISBN: (纸本)9798350330717

To improve the task scheduling efficiency in computing power networks, this paper proposes a global task scheduling method based on network measurement and prediction in computing power networks (GTS-MP), which selects a more suitable computing power cluster and storage platform with a higher network availability score between them to implement tasks. The experiments are conducted on a test bench of urban area networks in 12 cities. The results show that this method greatly reduces the execution time of artificial intelligence model training tasks, with a maximum reduction of 40% compared to other plans. © 2023 IEEE.

关键词： computing power network global scheduling network availability score task offloading

来源：评论

学校读者我要写书评

暂无评论

Self-Paced Self-Training Based on K-Nearest Neighbors 4

Self-Paced Self-Training Based on K-Nearest Neighbors

引用

4th International Conference on Machine Learning and computer Application, ICMLCA 2023

作者： Ma, Yongzheng Sun, Zhengkai Zhang, Yintong Guan, Hongjiao Qiao, Xinxiao Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center National Supercomputer Center in Jinan Qilu University of Technology Shandong Academy of Sciences Jinan China Shandong Provincial Key Laboratory of Computer Networks Shandong Fundamental Research Center for Computer Science Jinan China

ISBN: (纸本)9798400709449

Self-training is a classical semi-supervised classification method, which iteratively trains labeled samples that are enhanced using unlabeled samples. However, traditional self-training selects pseudo-labeled examples without distinguishing between different classes. Furthermore, wrong pseudo-labeled samples accumulate, leading to an increase in learning bias. In this paper, we propose a self-paced self-training based on K-nearest neighbors (SPST-KNN). SPST-KNN is composed of three main components: sample difficulty calculation, sample type determination, and self-paced selection based on KNN. The types of examples, including safe, boundary, and outlying examples, are determined by KNN. According to the principle of self-paced learning, SPST-KNN preferentially selects easy pseudo-labeled examples in each sample type to be added into the labeled sample set. Experiments and ablation study on 20 real-world datasets have demonstrated the superiority and effectiveness of the proposed SPST-KNN. © 2023 ACM.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Research on Low-Light Image Enhancement method based on UNet 5

Research on Low-Light Image Enhancement method based on UNet

引用

5th International Conference on Telecommunications, Optics, and computer Science, TOCS 2024

作者： Zhu, Fuyu Fei, Rong Ji, Chengjie School of Computer Science and Engineering Xi'an University of Technology Xi'an China Shannxi Key Laboratory for Network Computing and Security Technology Xi'an University of Technology Xi'an China

ISBN: (纸本)9781510691667

As technology develops continuously, digital images are being applied more and more extensively in daily life. Digital images are mainly divided into two categories: one is the images obtained by photosensitive devices such as cameras, and the other is the images generated by software. Nevertheless, in low - light conditions, the images captured by traditional photosensitive devices typically suffer from issues like low luminance and subpar visual qualities. These problems pose significant challenges to subsequent image processing and practical applications. Thus, the technology of enhancing low - light images has emerged as a crucial research area for the purpose of elevating image quality. This paper puts forward a solution centered around the improved UNet network to address the issue of image enhancement in low - light settings. The UNet network effectively performs image enhancement through feature extraction, feature fusion, and restoration operations. However, the classical UNet can only process single - channel grayscale images, and the loss function it uses has relatively limited effects in low - light image processing. Therefore, this paper adjusts the number of channels of the UNet network so that it can process three - channel RGB images and improves the loss function. We compared the enhancement effects of L1 Loss, VGG Loss, and SmoothL1 Loss respectively, and carried out a quantitative analysis through evaluation metrics such as PSNR and SSIM. Ultimately, through the combination of the three loss functions and their weighted optimization, the enhancement effect of low - light images is notably enhanced. The innovation of this paper lies in expanding the structure of the UNet network to make it adaptable to three - channel image processing, and improving the quality of image enhancement by improving the loss function. This approach is capable of efficiently enhancing the visual impact of low - light images, thereby offering a practical solution for imag

关键词： Photointerpretation

来源：评论

学校读者我要写书评

暂无评论

General Simultaneous Localization and Mapping Scheme for mmWave Communication Systems

引用

IEEE Internet of Things Journal 2024年第12期11卷 22521-22536页

作者： Du, Tao Yang, Jie Wen, Chao-Kai Xia, Shuqiang Jin, Shi Southeast University National Mobile Communications Research Laboratory Nanjing210096 China Southeast University Key Laboratory Of Measurement And Control Of Complex Systems Of Engineering Ministry Of Education The Frontiers Science Center For Mobile Information Communication And Security Nanjing210096 China Institute Of Communications Engineering National Sun Yat-sen University Kaohsiung80424 Taiwan Zte Corporation The State Key Laboratory Of Mobile Network And Mobile Multimedia Technology Shenzhen518055 China Southeast University National Mobile Communications Research Laboratory The Frontiers Science Center For Mobile Information Communication And Security Nanjing210096 China

Utilizing high-resolution antenna arrays and wide bandwidth of the millimeter-wave (mmWave) spectrum in 5G new radio (NR) mmWave communication systems holds the potential for high-throughput data transmission while enabling user localization and environmental mapping. However, the majority of existing simultaneous localization and mapping (SLAM) algorithms rely on methods akin to the extended Kalman filter for generating initial map features. These methods prove ineffective when the measurement dimension is insufficient. Furthermore, there is a notable absence of research exploring mmWave prototype systems to evaluate and compare the performance and viability of various SLAM algorithms. To address these challenges, we propose an innovative probability hypothesis density (PHD) generation scheme for birth events and have developed a prototype system. Our approach, referred to as PHD-SLAM, exhibits remarkable effectiveness even in scenarios where the measurement dimension falls short of map features. This means it can function seamlessly with only delay or angle information available. Additionally, we have designed a 28GHz mmWave beam scanning prototype system that leverages the 5G NR frame for accomplishing SLAM algorithms. Following this, we conducted extensive simulations and experimental evaluations to gauge the performance of several leading-edge SLAM algorithms under diverse mmWave circumstances, encompassing PHD-based and belief propagation (BP) SLAM algorithms. Our analysis reveals that both PHD and BP SLAM can achieve agent localization precision within a decimeter and mapping precision within a meter, capitalizing on the angle parameters of mmWave signals. While BP SLAM showcases reduced computational demand, its estimation precision is marginally inferior to that of PHD-SLAM. © 2014 IEEE.

关键词： Belief propagation

来源：评论

学校读者我要写书评

暂无评论

Research on Classroom Behavior Target Detection Algorithm based on YOLOv7 6

Research on Classroom Behavior Target Detection Algorithm ba...

引用

6th IEEE Advanced information Management, Communicates, Electronic and Automation Control Conference, IMCEC 2024

作者： Qiu, Wenke Wu, Xiaoming Liu, Xiangzhi Computer Science and Qilu University of Technology Shandong Academy of Sciences ShanDong Jinan China Qilu University of Technology Shandong Academy of Sciences Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center National Supercomputer Center in Jinan Jinan China Shandong Fundamental Research Center for Computer Science Shandong Provincial Key Laboratory of Computer Networks Jinan China

ISBN: (纸本)9798350316520

The integration of video information for the purpose of detecting students' behavior in the classroom, with the aim of enhancing teaching efficiency and maintaining a stable classroom environment, offers significant assistance. However, existing models usually struggle to achieve good accuracy due to the lack of datasets, dense targets, interclass occlusion, and other issues. To address these difficulties, the CD-YOLO classroom behavior detection algorithm is proposed in this paper. And the classroom behavior dataset (DSL-Dataset) is created. We propose a loss function named VAIoU, which can be combined with vector ratio to filter the prediction box. To further improve the detection accuracy of occluded targets, we propose an attention module CM module with command feature fusion. Moreover, in this paper, we improve the convolutional kernel to design a lightweight convolutional kernel FFconv. Experimental results on DSL-Dataset show that CD-YOLO outperforms existing one-stage detection models in terms of detection accuracy. © 2024 IEEE.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

A two-path fiber-optic AE demodulation algorithm based on Numerically Stable Direct Least Squares Fitting of Ellipses

引用

Journal of Lightwave Technology 2025年

作者： Li, Xinhao Zhang, Faxiang Sun, Zhihui Jiang, Shaodong Jinan China Shandong Fundamental Research Center for Computer Science Shandong Provincial Key Laboratory of Industrial Network and Information System Security Jinan China

The 3 × 3 coupler demodulation algorithm commonly employs ellipse fitting algorithms (EFAs) to eliminate coupler asymmetry. In the detection scenario of fiber-optic acoustic emission (AE) sensor arrays, the amount of data increases significantly due to the high sampling rate and multi-channel characteristics. Traditional EFAs require substantial computational resources and involve complex calculations, which severely hinder demodulation efficiency. To address these issues, We propose a two-path demodulation algorithm based on Numerically Stable Direct Least Squares Fitting of Ellipses (NSDLS). The algorithm only two input signals and divides the fitting process into linear and quadratic components, reducing computational complexity. The triggered ellipse fitting judgment method is introduced to mitigate background noise impact on fitting results and enhance demodulation efficiency. Experimental results demonstrate that the proposed two-path demodulation algorithm based on NSDLS reduces time consumption by approximately 40 times and the CPU utilization by 40% compared with traditional EFAs demodulation algorithm, improves SINAD by 18.93 dB compared with the improved Naval Postgraduate School (NPS) method. The algorithm achieves a dynamic range of 126.04 dB @ 1 kHz, 0.075% THD, and 99.999% response linearity. This method maintains high demodulation accuracy while significantly reducing computational resource demands, lowering hardware costs, and improving overall demodulation efficiency. © 1983-2012 IEEE.

关键词： Acoustic emission testing

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：