检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Sheng, Kaicheng Xue, Junxiao Zhang, Hui Research Center for Space Computing System Zhejiang Lab Hangzhou China College of Civil and Environmental Engineering Carnegie Mellon University Pittsburgh United States

The increasing availability of high-resolution satellite imagery has created immense opportunities for various applications. However, processing and analyzing such vast amounts of data in a timely and accurate manner poses significant challenges. The paper presents a new satellite image processing architecture combining edge and cloud computing to better identify man-made structures against natural landscapes. By employing lightweight models at the edge, the system initially identifies potential man-made structures from satellite imagery. These identified images are then transmitted to the cloud, where a more complex model refines the classification, determining specific types of structures. The primary focus is on the trade-off between latency and accuracy, as efficient models often sacrifice accuracy. We compare this hybrid edge-cloud approach against traditional"bent-pipe" method in virtual environment experiments as well as introduce a practical model and compare its performance with existing lightweight models for edge deployment, focusing on accuracy and latency. The results demonstrate that the edge-cloud collaborative model not only reduces overall latency due to minimized data transmission but also maintains high accuracy, offering substantial improvements over traditional approaches under this scenario. © 2024, CC0.

关键词： Satellite imagery

来源：评论

学校读者我要写书评

暂无评论

A Neural Network Parallel computing Approach for Arithmetic Test Loads 13

A Neural Network Parallel Computing Approach for Arithmetic ...

引用

13th International Conference of Information and Communication Technology, ICTech 2024

作者： Wang, HongLiang Xing, GuoYang Wang, JunLi Gao, Xing Li, XueRao Liu, XingYu Shenyang Institute of Computing Technology Chinese Academy of Sciences Shenyang China University of Chinese Academy of Sciences Beijing China Liaoning Province Human-Computer Interaction System Engineering Research Center Based on Digital Twin Shenyang China Affiliated Hospital of Guangdong Medical University Guangzhou China Shengjing Hospital of China Medical University Shenyang China

ISBN: (纸本)9798350376258

The development of server arithmetic benchmarking is of great significance to the development of the whole chip and the whole server arithmetic. Currently, the existing arithmetic tests are categorized into traditional server arithmetic tests, e.g., Linpack, and AI server arithmetic tests, e.g., MlPerl. Traditional server arithmetic tests are insufficient to reflect the arithmetic power of AI servers, and existing AI servers ignore the limitations of the parallelism method, which leads to the instability of the final test results. In this paper, we propose a parallel computing method, DPipe-C, for benchmarking to balance the contradiction between long testing time and unstable test results during training tests. The method uses a dual-pipe approach in order to reduce the size of bubbles in the parallel process and improve the overall throughput of the system. It also uses synchronization to ensure the stability of its results. In the different stages of data transmission, the data is compressed to improve the overall training efficiency, and the delayed delivery error method is used to reduce the impact of the compression process on the training results to ensure the stability of the results. Using this method, the overall throughput of the system is increased by nearly 20%. At the same time, the training accuracy can reach the original accuracy of the model. ©2024 IEEE.

关键词： Parallel processing systems

来源：评论

学校读者我要写书评

暂无评论

Security and privacy with opacity-based state observation for finite state machine

Security and privacy with opacity-based state observation fo...

引用

作者： Zhang, Zhipeng Xia, Chengyi Chen, Shengyong Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology Tianjin University of Technology Tianjin China Engineering Research Center of Learning-Based Intelligent System Ministry of Education Tianjin University of Technology Tianjin China

Topics on security and privacy are very fundamental and important in the Internet of Things (IoT). In this paper, we explore the impact of state observation on the opacity of IoT, which is modeled by Moore-type finite state machine. Firstly, to characterize the intrusion estimator under the event-state observation, a multi-linear algebraic expression is proposed through the Boolean semi-tensor product (STP), and on the basis of the novel representation, the validation criterion is given by some matrix manipulations. Subsequently, considering that the input information may not be measured due to the physical shielding of the system against external intruders, we continue to discuss the opacity with state observation by a permutation structure matrix. Meanwhile, the verification of opacity with event-state observation is generalized to the one with only state observation. Finally, we offer an example to illustrate the effectiveness of the newly proposed method. Taken together, the current results are helpful to provide a novel theoretical approach for investigating the privacy-preserving problem in the area of IoT and complex cyber-physical networks. © 2021 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Deep User and Item Inter-matching Network for CTR Prediction 28th

Deep User and Item Inter-matching Network for CTR Predicti...

引用

28th International Conference on Database systems for Advanced Applications, DASFAA 2023

作者： Yuan, Zhiyang Xiao, Yingyuan Yang, Peilin Hao, Qingbo Wang, Hongya Engineering Research Center of Learning-Based Intelligent System Ministry of Education Tianjin University of Technology Tianjin300384 China Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology Tianjin University of Technology Tianjin300384 China College of Computer Science and Engineering Donghua University Shanghai China

ISBN: (纸本)9783031306716

CTR prediction plays an important role in increasing company revenue and user experience, and many efforts start with historical behavior to uncover user interest. There are two main problems with previous works: (1) When most previous works mined interests from users’ historical behaviors, they only focus on implicit or explicit interests. (2) When most previous works mined user interests through the relationship between target users and similar users, the representation of target users was not rich and accurate enough, resulting in less accurate item recommendations. Therefore the Deep User and Item Inter-Matching Network (DUIIN) is proposed in this paper to solve the above problems. First, we design Item-to-Item Network (IIN), using two different sub-networks Evolving Interest Network (EIN) and Feature Interaction Network (FIN) to mine the hidden interests and explicit interests shown by users’ historical behaviors, respectively. Then the User-to-User Network (UUN) is designed to mine user interests through the relationship between target users and similar users after representing the target users more accurately and richly. The experimental results show that the DUIIN model proposed in this paper performs better than other state-of-the-art models. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

LEMTL: Enhancing the pharmacokinetic predictions of multitask learning with existing pharmacokinetics

LEMTL: Enhancing the pharmacokinetic predictions of multitas...

引用

IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

作者： Ying Li Xiao Deng Jingsong Lv Hongyang Chen Yao Yang Research Center for Space Computing System Zhejiang Lab Hangzhou P. R. China Research Center for Data Hub and Security Zhejiang Lab Hangzhou P. R. China

ISBN: (数字)9798350386226

ISBN: (纸本)9798350386233

Evaluation of pharmacokinetic properties such as absorption, distribution, metabolism, excretion, and toxicity is essential but costly and time-consuming in the research and development of new drug candidates. This challenge has driven many researchers to use multitask learning for in silico predictions. Given extra information such as SMILES (simplified molecular input line entry system) as input data, existing multitask learning methods can usually have a good performance in predicting pharmacokinetics, which are served as labels/tasks. However, most of these methods are based solely on additional information, overlooking potentially available pharmacokinetics, which can be utilized to further enhance the prediction of unavailable ones. To address this limitation, we propose Label Enhanced MultiTask Learning (LEMTL), a framework to improve the prediction accuracy of unknown pharmacokinetics of existing multitask learning methods by aggregating shared feature representations learned from additional information and from known pharmacokinetics. Experiments on synthetic and real datasets demonstrate significant improvements over the associated multitask learning approaches when incorporated within the LEMTL framework.

关键词： Learning systems Drugs Toxicology Accuracy Absorption Feature extraction Pharmacokinetics Metabolism Bioinformatics research and development

来源：评论

学校读者我要写书评

暂无评论

Beamforming Design of ISAC system for Enhancing Sensing Distance

Beamforming Design of ISAC System for Enhancing Sensing Dist...

引用

Future Communications and Networks (FCN), International Conference on

作者： Yuan Liu Ruihong Jiang Yongdong Zhu Bin Yang Shuyuan Zhao Research Center for Space Computing System Zhejiang Laboratory Hangzhou China State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing China

ISBN: (数字)9798331512880

ISBN: (纸本)9798331512897

In this paper, a dual-functionality base station (BS) integrated sensing and communication (ISAC) is exploited to sense one intended target and communicate with multiple users simultaneously. Meanwhile, rate splitting multiple access (RSMA) technology is used to improve communication quality. We formalize a problem aimed at maximizing sensing distance by optimizing the beamforming design of the BS and the allocation of common rate portions, where the signal to interference plus noise ratio (SINR) gain constraint of sensing echo signal and the individual communication rate requirements of each user are satisfied. Due to the non-convexity of the formulated problem, a successive convex approximation (SCA)-based algorithm is proposed to tackle it. Simulation results manifest that as the transmit power increases, the achievable sensing distance increases, and as the SINR threshold of the sensing echo signal increases, the achievable sensing distance decrease. Furthermore, it indicates that the RSMA-based protocol offers superior sensing distance gain compared to the SDMA-based protocol. This performance advantage becomes more pronounced when the communication rate threshold is high, resulting in a relatively large performance gap between the two protocols.

关键词： Base stations Protocols Array signal processing Simulation Interference Integrated sensing and communication Approximation algorithms Resource management Iterative methods Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

EagerLog: Active Learning Enhanced Retrieval Augmented Generation for Log-based Anomaly Detection

EagerLog: Active Learning Enhanced Retrieval Augmented Gener...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Duan, Chiming Jia, Tong Yang, Yong Liu, Guiyang Liu, Jinbu Zhang, Huxing Zhou, Qi Li, Ying Huang, Gang School of Software and Microelectronics Peking University Beijing China Institute for Artificial Intelligence Peking University Beijing China National Engineering Research Center for Software Engineering Peking University Beijing China National Key Laboratory of Data Space Technology and System Beijing China Alibaba Cloud Computing Company Hangzhou China

ISBN: (纸本)9798350368741

Logs record essential information about system operations and serve as a critical source for anomaly detection, which has generated growing research interest. Utilizing large language models (LLMs) within a retrieval-augmented generation (RAG) framework for log-based anomaly detection is an effective approach due to its strong generalization capabilities and efficient few-shot performance. However, the effectiveness of this method hinges on the quality of the knowledge source, which can be impacted by noise and changes within the software systems. Facing these problems, in this paper, we propose a novel log-based anomaly detection method named EagerLog, employing active learning to choose the logs for humans to label, thereby adding them to the knowledge source, thus enhancing the knowledge source and maintaining its quality. Our experiments on three open datasets (BGL, Thunderbird, Zookeeper) and one industrial dataset demonstrate that EagerLog can achieve 93.65% F1 score with approximately 10 labeled log sequences, surpassing existing methods by 15.32%. © 2025 IEEE.

关键词： Active Learning Anomaly Detection Log Analysis Retrieval Augmented Generation

来源：评论

学校读者我要写书评

暂无评论

Automatic detection and classification algorithm of thyroid nodules in CT images based on YOLOv5s 9

Automatic detection and classification algorithm of thyroid ...

引用

9th International Conference on Computer and Communications, ICCC 2023

作者： Guo, Ruifeng Xu, Lei Guo, Jiahui Wang, Hongliang Shenyang Institute of Computing Technology Chinese Academy of Sciences Shenyang China University of Chinese Academy of Sciences Shenyang China Liaoning Prov. Human-Computer Interaction System Engineering Research Center Based on Digital Twin Shenyang China The First Hospital of China Medical University Department of Endocrinology and Metabolism Shenyang China

ISBN: (纸本)9798350317251

Thyroid nodular lesions are one of the most common diseases in the population. As the number of patients increases, it is clinically important to provide physicians with an accurate early aid to determine the benignity and malignancy of thyroid nodules in order to avoid unnecessary surgery and reduce patient burden. In this paper, we propose an improved Yolov5s-based target detection algorithm for benign and malignant thyroid nodules for thyroid nodule classification and identification. Replacing the backbone network of Yolov5s with ResNet18 to reduce the complexity of the model and the number of parameters;introducing the CBAM attention mechanism into the Yolov5s algorithm to improve the ability to extract features;Modify the spatial pyramid pooling module SPP to SimSPPF to improve the training speed of the algorithm;replace GIoU Loss with CIoU-Focal Loss as the loss function of the algorithm to improve the localization accuracy while increasing the regression rate of the bounding box. The experimental results show that the improved Yolov5s achieved an average accuracy of 92.9%, which is 2.9 percentage points higher than the average accuracy of the original algorithm, and can effectively improve the accuracy and efficiency of nodule detection. © 2023 IEEE.

关键词： Computerized tomography

来源：评论

学校读者我要写书评

暂无评论

research on Sensitive Text Classification based on Knowledge Base and Hybrid Network 8

Research on Sensitive Text Classification Based on Knowledge...

引用

8th IEEE International Conference on Computer and Communications, ICCC 2022

作者： Li, Yongfeng Wang, Hongliang Li, Xueting Xiu, Pengfei Shenyang Institute of Computing Technology Chinese Academy of Sciences Shenyang China University of Chinese Academy of Sciences Beijing China Liaoning Prov. Human-Computer Interaction System Engineering Research Center Based on Digital Twin Shenyang China

ISBN: (纸本)9781665450515

Aiming at the problems that convolution neural network is not sufficient to extract text features, it is difficult to capture long text structure information and sentence semantic relationship, and neural network can only extract the surface features of text, and it is difficult to obtain the implicit features of sentences when classifying text, this paper proposes a hybrid neural network text classification method based on a knowledge base. By extracting keywords from the category text, the keyword weight is calculated according to the co-occurrence relationship, and the category keyword knowledge base is constructed. It integrates CNN and bilstm, introduces attention mechanism, enhances the ability to extract local and temporal features of text, and improves the accuracy of text classification in combination with knowledge base information. The experimental results on the CNews dataset show that compared with TextCNN, TextRCNN, BILSTM and Bert models, this model can effectively improve the accuracy of text classification. © 2022 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

research on Load Forecasting based on IPSO-BILSTM-GRU Model 8

Research on Load Forecasting Based on IPSO-BILSTM-GRU Model

引用

8th IEEE International Conference on Computer and Communications, ICCC 2022

作者： Li, Xueting Wang, Hongliang Xiu, Pengfei Li, Yongfeng Shenyang Institute of Computing Technology Chinese Academy of Sciences Shenyang China University of Chinese Academy of Sciences Beijing China Liaoning Prov. Human-Computer Interaction System Engineering Research Center Based on Digital Twin Shenyang China

ISBN: (纸本)9781665450515

High accuracy of resource load prediction can provide basis for real-time task scheduling, thus reducing energy consumption. Aiming at the problems that the CPU and memory load changes in cloud computing platform cannot be accurately predicted and it is difficult to adjust the parameters of common timing prediction algorithms, a combined prediction model of bidirectional long short-term memory network (BILSTM) and gated recurrent unit (GRU) based on improved particle swarm optimization (IPSO), IPSO-BILSTM-GRU, was proposed. Our model combines the characteristics of high prediction accuracy of BILSTM network and short prediction time of GRU network. The IPSO algorithm is used to optimize the model parameters. In the experiment, the combined values of CPU and memory indexes of Google data center are selected as input data and compared with many typical time series prediction algorithms. Experimental results show that the MAPE (mean absolute percentage error) of IPSO-BILSTM-GRU model is reduced by 1.0% to 4.4%, and the training time is shortened by more than 11%. The experimental results verify that the method has higher accuracy and strong applicability in time series prediction, which provides an important scientific basis for elastic capacity expansion using load prediction results. © 2022 IEEE.

关键词： Cloud computing

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：