检索结果-内蒙古大学图书馆

23rd International Conference on Autonomous Agents and Multiagent systems, AAMAS 2024

作者： Wu, Xiaoqiang Zhu, Qingling Lin, Qiuzhen Chen, Weineng Li, Jianqiang College of Computer Science and Software Engineering Shenzhen University Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China School of Computer Science and Engineering South China University of Technology Guangzhou China

Evolutionary reinforcement learning algorithms (ERLs), which combine evolutionary algorithms (EAs) with reinforcement learning (RL), have demonstrated significant success in enhancing RL performance. However, most ERLs rely heavily on Gaussian mutation operators to generate new individuals. When the standard deviation is too large or small, this approach will result in the production of poor or highly similar offspring. Such outcomes can be detrimental to the learning process of the RL agent, as too many poor or similar experiences are generated by these individuals. In order to alleviate these issues, this paper proposes an Adaptive Evolutionary Reinforcement Learning (AERL) method that adaptively adjusts both the standard deviation and the evaluation process. By tracking the performance of new individuals, AERL maintains the mutation strength within a suitable range without the need for additional gradient computations. Moreover, the proposed AERL approach early terminates unnecessary evaluations and discards experiences arising from poor individuals, thereby resulting in enhanced learning efficiency. Empirical assessments conducted on a variety of continuous control problems demonstrate the effectiveness of the AERL method. © 2024 International Foundation for Autonomous Agents and Multiagent systems.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

FDGLib: A Communication Library for Efficient Large-Scale Graph Processing in FPGA-Accelerated Data Centers

引用

Journal of computer Science & technology 2021年第5期36卷 1051-1070页

作者： Yu-Wei Wu Qing-Gang Wang Long Zheng Xiao-Fei Liao Hai Jin Wen-Bin Jiang Ran Zheng Kan Hu National Engineering Research Center for Big Data Technology and System School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Services Computing Technology and System Laboratory School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Cluster and Grid Computing Laboratory School of Computer Science and TechnologyHuazhong University of Science and TechnologyWuhan 430074China

With the rapid growth of real-world graphs,the size of which can easily exceed the on-chip(board)storage capacity of an accelerator,processing large-scale graphs on a single Field Programmable Gate Array(FPGA)becomes *** multi-FPGA acceleration is of great necessity and *** cloud providers(e.g.,Amazon,Microsoft,and Baidu)now expose FPGAs to users in their data centers,providing opportunities to accelerate large-scale graph *** this paper,we present a communication library,called FDGLib,which can easily scale out any existing single FPGA-based graph accelerator to a distributed version in a data center,with minimal hardware engineering *** provides six APIs that can be easily used and integrated into any FPGA-based graph accelerator with only a few lines of code *** the torus-based FPGA interconnection in data centers,FDGLib also improves communication efficiency using simple yet effective torus-friendly graph partition and placement *** interface FDGLib into AccuGraph,a state-of-the-art graph *** results on a 32-node Microsoft Catapult-like data center show that the distributed AccuGraph can be 2.32x and 4.77x faster than a state-of-the-art distributed FPGA-based graph accelerator ForeGraph and a distributed CPU-based graph system Gemini,with better scalability.

关键词： data center accelerator graph processing distributed architecture communication optimization

来源：评论

学校读者我要写书评

暂无评论

MAML MOT: Multiple Object Tracking Based on Meta-Learning

MAML MOT: Multiple Object Tracking Based on Meta-Learning

引用

2024 IEEE International Conference on systems, Man, and Cybernetics, SMC 2024

作者： Chen, Jiayi Deng, Chunhua School of Computer Science and Technology Wuhan University of Science and Technology Wuhan430065 China Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System Wuhan430065 China

ISBN: (纸本)9781665410205

With the advancement of video analysis technology, the multi-object tracking (MOT) problem in complex scenes involving pedestrians is gaining increasing importance. This challenge primarily involves two key tasks: pedestrian detection and re-identification. While significant progress has been achieved in pedestrian detection tasks in recent years, enhancing the effectiveness of re-identification tasks remains a persistent challenge. This difficulty arises from the large total number of pedestrian samples in multi-object tracking datasets and the scarcity of individual instance samples. Motivated by recent rapid advancements in meta-learning techniques, we introduce MAML MOT, a meta-learning-based training approach for multi-object tracking. This approach leverages the rapid learning capability of meta-learning to tackle the issue of sample scarcity in pedestrian re-identification tasks, aiming to improve the model's generalization performance and robustness. Experimental results demonstrate that the proposed method achieves high accuracy on mainstream datasets in the MOT Challenge. This offers new perspectives and solutions for research in the field of pedestrian multi-object tracking. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

SSD object detection algorithm based on knowledge map

引用

International Journal of Wireless and Mobile Computing 2024年第2期27卷 152-160页

作者： Huang, Li Wang, Xiaofeng Lu, Jianhua Hu, Wei Zhang, Changrong College of Computer Science and Technology Wuhan University of Science and Technology Hubei Wuhan China China Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System Wuhan University of Science and Technology Hubei Wuhan China Key Laboratory of Metallurgical Equipment and Control Technology of the Ministry of Education Wuhan University of Science and Technology Hubei Wuhan China Precision Manufacturing Research Institute Research Centre for Biomimetic Robot and Intelligent Measurement and Control Wuhan University of Science and Technology Hubei Wuhan China

With the pervasive integration of artificial intelligence into all aspects of human life, talent emerges as a primary resource. Upon analysing the current state of talent training in higher education institutions, issues such as dispersed knowledge points, overlapping content, a singular practice approach and ineffective evaluation have been identified. In response to these challenges, this paper proposes a multidisciplinary and comprehensive practical teaching methodology grounded in the knowledge graph framework. It delves into diverse paths for practical teaching and assessment, including aspects like teaching objectives, problem decomposition, resource integration, implementation methods and performance evaluation. The practical application of the SSD algorithm in researching service robot indoor object detection serves as an illustrative example. Employing the holistic practical teaching approach facilitated by the knowledge graph, this model guides students in acquiring object detection expertise, thereby enhancing their comprehensive development. Copyright © 2024 Inderscience Enterprises Ltd.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

An Anomaly Detection Model for Log Analysis Based on Multi-Head Self-Attention Mechanism 8

An Anomaly Detection Model for Log Analysis Based on Multi-H...

引用

8th International Conference on Computational Intelligence and Applications, ICCIA 2023

作者： Jiao, Yihan Fu, Fa Liu, Zhaojun Zeng, Dechao Huang, Liangliang Yuan, Quan Li, Liang Hainan University College of Computer Science and Technology Hainan China Key Laboratory of PK System Technologies Research of Hainan Hainan China

ISBN: (纸本)9798350339741

Current research on log analysis is mainly implemented on complex recurrent or convolutional neural networks. In this study, we propose a new concise network architecture with an excellent performance which based on multi-head self-attention mechanism. Our model, named as MALAD, requires significantly less number of parameters and perfectly adapting to the high-speed development of massively parallel computing. MALAD learns the way of positional encoding autonomously, so it is able to obtain the most suitable approach to understand the positional information from each kind of log data from different system. In the process of experiments, our model achieves well performance on HDFS and BGL datasets [1]. This research can be extended to solve other tasks of log analysis such as performance modeling and system failure diagnosis. © 2023 IEEE.

关键词： Network architecture

来源：评论

学校读者我要写书评

暂无评论

Spectral period doubling and encoding of dissipative optical solitons via gain control

引用

PhotoniX 2024年第1期5卷 290-305页

作者： Kangwen Yang Yi Zhou Yuqing Ling Kevin K.Tsia Heping Zeng Kenneth K.Y.Wong Shanghai Key Laboratory of Modern Optical System and Engineering Research Center of Optical Instrument and SystemMinistry of EducationSchool of Optical Electrical and Computer EngineeringUniversity of Shanghai for Science and TechnologyShanghai 200093China Department of Electrical and Electronic Engineering The University of Hong KongHong KongChina Advanced Biomedical Instrumentation Centre Hong Kong Science ParkHong KongChina State Key Laboratory of Precision Spectroscopy East China Normal UniversityShanghai 200062China

Period-doubling bifurcation,as an intermediate state between order and chaos,is ubiquitous in all disciplines of nonlinear ***,previous experimental observations of period doubling in ultrafast fiber lasers are mainly restricted to self-sustained steady state,controllable manipulation and dynamic switching between period doubling and other intriguing dynamical states are still largely ***,we propose to expand the vision of dissipative soliton periodic doubling,which we illustrate experimentally by reporting original spontaneous,collisional,and controllable spectral period doubling in a polarization-maintaining ultrafast fiber ***,the spontaneous period doubling can be observed in both single-and *** mechanism of the switchable state and periodic doubling was revealed by numerical ***,state transformation of individual solitons can be resolved during the collision of triple solitons involving stationary,oscillating,and period ***,controllable deterministic switching between period doubling and other dynamical states,as well as exemplifying the application of period-doubling-based digital encoding,is achieved under programmable pump *** results open a new window for unveiling complex Hopf bifurcation in dissipative systems and bring useful insights into nonlinear science and applications.

关键词： Period doubling Mode locking Ultrafast fiber lasers

来源：评论

学校读者我要写书评

暂无评论

On the Security of Smart Home systems:A Survey

引用

Journal of computer Science & technology 2023年第2期38卷 228-247页

作者：袁斌万俊吴宇晗邹德清金海 School of Cyber Science and Engineering Huazhong University of Science and TechnologyWuhan 430074China Hubei Key Laboratory of Distributed System Security Huazhong University of Science and Technology Wuhan 430074China Hubei Engineering Research Center on Big Data Security Huazhong University of Science and Technology Wuhan 430074China National Engineering Research Center for Big Data Technology and System Huazhong University of Science and TechnologyWuhan 430074China Services Computing Technology and System Lab Huazhong University of Science and TechnologyWuhan 430074China Shenzhen Huazhong University of Science and Technology Research Institute Shenzhen 518057China School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Cluster and Grid Computing Lab Huazhong University of Science and TechnologyWuhan 430074China

Among the plethora of IoT(Internet of Things)applications,the smart home is one of the ***,the rapid development of the smart home has also made smart home systems a target for ***,researchers have made many efforts to investigate and enhance the security of smart home *** a more secure smart home ecosystem,we present a detailed literature review on the security of smart home ***,we categorize smart home systems’security issues into the platform,device,and communication *** exploring the research and specific issues in each of these security areas,we summarize the root causes of the security flaws in today's smart home systems,which include the heterogeneity of internal components of the systems,vendors'customization,the lack of clear responsibility boundaries and the absence of standard security ***,to better understand the security of smart home systems and potentially provide better protection for smart home systems,we propose research directions,including automated vulnerability mining,vigorous security checking,and data-driven security analysis.

关键词： IoT(Internet of Things) smart home IoT security smart home security

来源：评论

学校读者我要写书评

暂无评论

PATH OPTIMIZATION FOR DYNAMIC OBSTACLE AVOIDANCE OF PIONEER ROBOT BASED ON DDPG SMART AGENT

引用

UPB Scientific Bulletin, Series D: Mechanical Engineering 2023年第2期85卷 29-42页

作者： Zheng, Qingchun Peng, Zhi Zhu, Peihao Zhao, Yangyang Ma, Wenpeng Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control Tianjin University of Technology Tianjin300384 China China School of Mechanical Engineering Tianjin University of Technology Tianjin300384 China School of Computer Science and Engineering Tianjin University of Technology Tianjin300384 China

To improve the path efficiency for dynamic obstacle avoidance algorithms of Pioneer robots, this paper proposes an approach to create deep deterministic policy gradient (DDPG) smart agents with recurrent neural networks (RNN) to optimize the path strategy. This approach is realized by training RNN with the backpropagation (BP) algorithm to train the agent. First, the SolidWorks model and obstacle avoidance model of the Pioneer robot is established. The LIDAR sensor is used to obtain environmental information. Further, deep reinforcement learning obstacle avoidance algorithm based on RNN is presented. The dynamic obstacle avoidance path of the Pioneer robot is further simulated in two different virtual robot experiment platform (V-REP) scenarios. The simulation results show that the proposed algorithm shortens the training time by 26.3% compared with the DDPG algorithm. Our proposed scheme can improve the path efficiency for dynamic obstacle avoidance of the Pioneer robot. © 2023, Politechnica University of Bucharest. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

PieBridge: Fast and Parameter-Efficient On-Device Training via Proxy Networks 24

PieBridge: Fast and Parameter-Efficient On-Device Training v...

引用

22nd ACM Conference on Embedded Networked Sensor systems, SenSys 2024

作者： Yin, Wangsong Xu, Daliang Huang, Gang Zhang, Ying Wei, Shiyun Xu, Mengwei Liu, Xuanzhe School of Computer Science Peking University Beijing China National Key Laboratory of Data Space Technology and System Beijing China Zhongguancun Laboratory Beijing China Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9798400706974

On-device training Neural Networks (NNs) has been a crucial catalyst towards privacy-preserving and personalized mobile intelligence. Recently, a novel training paradigm, namely Parameter-Efficient Training (PET), is attracting attention in both the machine learning and system community. In our preliminary measurements, we find PET well-suited for on-device scenarios;yet, its parameter efficiency does not translate coequal to time efficiency on resource-constrained devices, as the training time is dominated by the frozen *** this end, this work presents PieBridge, an on-device training framework with both time and parameter efficiency. Its key idea is to dynamically approximate the frozen layers to cheaper ones (subnets) with data awareness during PET To achieve effective and efficient approximate training, we introduce (1) a pre-training-assisted on-cloud subnets generation method and (2) an edge-friendly on-device data-aware subnets routing method. The subnets generation method performs fine-grained pruning and latent space alignment to generate a series of high-quality proxy subnets with varying speed-accuracy trade-offs for the deployment-ready NN. The subnets routing method perceives data diversity from two unique perspectives (referred to as importance and difficulty). The routing strategy is provided by an offline-learning and online-estimation fusion, which is accurate, end-to-end and cost-effective on devices. Through extensive experiments, we show that PieBridge exhibits up to 2.5X training speedup compared to state-of-the-art PET methods, and up to 6.6X speedup compared to traditional full model training and other on-device training frameworks, without compromising parameter efficiency and accuracy. © 2024 Copyright is held by the owner/author(s).

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

TERMS: Task management policies to achieve high performance for mixed workloads using surplus resources

引用

Journal of Parallel and Distributed Computing 2022年 170卷 74-85页

作者： Yu, Jinyu Tong, Wei Lv, Pengze Feng, Dan Wuhan National Laboratory for Optoelectronics Key Laboratory of Information Storage System Engineering Research Center of Data Storage Systems and Technology School of Computer Science and Technology Huazhong University of Science and Technology Ministry of Education of China Hubei Wuhan China

Resource contentions and performance interferences can lead to workload performance degradation in mixed-workload deployment clusters. Previous work guarantees the resource requirements of latency-sensitive tasks and reduces performance losses to batch jobs by reclaiming surplus resources from over-provisioned tasks. While the fragmentation of resources leads to a mismatch between provisioned resources and task requirements, resulting in high operation overheads and losses of task fairness. This paper proposes TERMS, the task management policies based on task relevance, resource distribution, and task fairness to achieve efficient and low-cost task management. TERMS mainly includes three types of management policies. The task scheduling policy can schedule new tasks according to task relevance. Task selection strategies select tasks for resource provisioning and task resumption based on resource requirements and task fairness. If necessary, the node selection strategy can be used to choose befitting target nodes based on task relevance and node resource information for task migration when eliminating straggler tasks. Evaluation results show that TERMS can further improve the performance of latency-sensitive services and batch jobs, reduce management overheads, and avoid operation failures. © 2022 Elsevier Inc.

关键词： Scheduling

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：