检索结果-南通市图书馆

面向博弈对抗的多智能体强化学习建模与迁移技术

在线全文

学校读者我要写书评

暂无评论

指挥与控制学报 2024年第2期10卷 226-231页

作者：李渊刘运韬徐新海万珂嘉军事科学院北京100190

多智能体博弈对抗问题涉及智能体之间的协作配合,传统基于博弈论等方法的解决方案并不适用于复杂场景的博弈对抗问题。基于强化学习的多智能体协同训练机制是近年来的研究热点。针对中国电子科技集团发布的多智能体博弈对抗问题,设计基... 详细信息

多智能体博弈对抗问题涉及智能体之间的协作配合,传统基于博弈论等方法的解决方案并不适用于复杂场景的博弈对抗问题。基于强化学习的多智能体协同训练机制是近年来的研究热点。针对中国电子科技集团发布的多智能体博弈对抗问题,设计基于值分解的多智能体深度强化学习方法,为每个智能体独立构建网络模型,通过引入混合网络连接各个智能体,训练时以混合网络指导各智能体网络更新,执行时各智能体网络独立运行,实现集中式学习、分散式执行的训练模式。针对同构异构场景,设计一种高效的迁移训练技术,提升多智能体强化学习方法在不同场景下的快速训练效率。对同构和异构博弈对抗问题分别进行测试,实验结果表明基于值分解的多智能体强化学习方法和迁移技术,能够有效提升智能体的协作行为以及训练效率。

关键词：多智能体对抗博弈深度强化学习迁移学习值分解混合网络训练效率

Experimental investigation on a novel liquid cooling device for a prismatic Li-ion battery module operating at high ambient temperature

在线全文

学校读者我要写书评

暂无评论

Science China(Technological Sciences) 2020年第10期63卷 2147-2153页

作者： xu JiaWei ZHOU TianShu xu xinhai School of Mechanical Engineering and Automation Harbin Institute of TechnologyShenzhen 518055China

A novel liquid cooling device for a prismatic LiFePO4 battery module was proposed and manufactured in this study in order to improve the thermal management performance of the battery module operating at high ambient temperature.A testing system was set up to experimentally measure temperatures in different locations of the battery module consisting of seven 60 Ah *** were conducted to investigate both the passive and active cooling performances of the proposed system at different ambient temperatures and discharging rates in regarding with the maximum temperature and difference between the maximum and minimum *** results clearly show that both the ambient temperature and discharging rate play important role on the maximum temperature of the battery *** cooling cannot meet the cooling requirement of the battery module particularly at high ambient temperature of 40℃.In contrary,liquid cooling can successfully reduce the maximum temperature to the required temperature range of the battery module even in high temperature environment and relatively high discharging *** effect of water inlet temperature on the cooling performance was also experimentally *** the inlet temperature of 28℃,the active cooling device can reduce the maximum temperature of the battery module to about 34.8℃after discharging at 0.6℃for 1000 *** temperature difference of only 3.8℃was also achieved which suggests a great uniform distribution of temperature in the battery module.

关键词： liquid cooling maximum temperature temperature difference Li-ion battery cold plate thermal management

Protocol-Aware Process Placement for MPI Programs

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2014年第4期23卷 701-705页

作者： xu xinhai LIN Yufei YANG xuejun State Key Laboratory of High Performance Computing National University of Defense Technology

Current parallel programs are often composed of lots of processes that communicate through message passing. The mapping of the program processes onto the underlying processor. By analyzing the MPI library,we find that the communication protocols have great influence upon the communication characteristics. The profiling information used by the existing process placement methods does not include the communication protocols. We propose and implement a method called Protocol-aware process placement(Pa PP) to incorporate the influences of communication protocols into the mapping problem. We demonstrate the validations of Pa PP by the experiments using NPB3.3-MPI benchmarks. The experimental results show that compared to basic process placement, the average execution time when using Pa PP scheme is reduced by 9.5%.

关键词： Process placement Communication protocol Graph mapping MPI

Simultaneous microwave characterization of wafer-level optoelectronic transceiver chips based on photonic sampling and mapping

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第2期67卷 337-338页

作者： Yutong HE xinhai ZOU Ying xu Zhihui LI Naidi CUI Junbo FENG Yali ZHANG Zhiyao ZHANG Shangjian ZHANG Yong LIU Ninghua ZHU Research Center for Microwave Photonics State Key Laboratory of Electronic Thin Films and Integrated DevicesSchool of Optoelectronic Science and EngineeringUniversity of Electronic Science and Technology of China Chongqing United Microelectronics Center (CUMEC) Xiongan Institute of Innovation Chinese Academy of Sciences

Photonic integrated circuits(PICs) promise future parallelism growths of high-performance communication, computation, and offer unprecedented bandwidth scalability with reduced power consumption as a viable replacement for an electrical wire with an optoelectronic transceiver consisting of a transmitter and a receiver integrated on the same wafer.

关键词：

Micro/nano flow chemistry by Beyond Limits Manufacturing

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Chinese Chemical Letters 2024年第5期35卷 3-12页

作者： Peiwen Liu Fang Zhao Jing Zhang Yunpeng Bai Jinxing Ye Bo Bao Xinggui Zhou Li Zhang Changlu Zhou xinhai Yu Peng Zuo Jianye Xia Lian Cen Yangyang Yang Guoyue Shi Lin xu Weiping Zhu Yufang xu xuhong Qian State Key Laboratory of Bioreactor Engineering Shanghai Key Laboratory of Chemical BiologySchool of PharmacyEast China University of Science and TechnologyShanghai 200237China State Key Laboratory of Chemical Engineering School of Chemical EngineeringEast China University of Science and TechnologyShanghai 200237China State Key Laboratory of Bioreactor Engineering Shanghai Collaborative Innovation Center for BiomanufacturingEast China University of Science and TechnologyShanghai 200237China School of Biomedical and Pharmaceutical Sciences Guangdong University of TechnologyGuangzhou 510006China School of Mechanical and Power Engineering East China University of Science and TechnologyShanghai 200237China Shanghai Key Laboratory of Green Chemistry and Chemical Processes Department of ChemistrySchool of Chemistry and Molecular EngineeringEast China Normal UniversityShanghai 200241China

In September 2018,we proposed the cutting-edge concept of“Beyond Limits Manufacturing”(BLM).BLM technology is based on the three-dimensional inner engraving or precise outer engraving of ultra-fast laser,to invent micro/nano scale flow chips or devices,which makes it possible for the microform,integration,economy,safety,high efficiency,green and intelligence of research,development and manufacturing process,so as to realize transformational manufacturing in the era of Industry *** this paper,we reviewed the representative results we made in the field of micro/nano flow chemistry during the implementation of the BLM major project(December 2019 to August 2023),and discussed its application prospects in micro/nano flow chemistry.

关键词： Beyond Limits Manufacturing Flow chemistry Pharmaceutical manufacturing Microreactor

基于组合训练的规则嵌入多智能体强化学习方法

在线全文

学校读者我要写书评

暂无评论

计算机应用研究 2022年第3期39卷 802-806页

作者：李渊徐新海军事科学院北京100190

多智能体强化学习方法在仿真模拟、游戏对抗、推荐系统等许多方面取得了突出的进展。然而,现实世界的复杂问题使得强化学习方法存在无效探索多、训练速度慢、学习能力难以持续提升等问题。该研究嵌入规则的多智能体强化学习技术,提出基... 详细信息

多智能体强化学习方法在仿真模拟、游戏对抗、推荐系统等许多方面取得了突出的进展。然而,现实世界的复杂问题使得强化学习方法存在无效探索多、训练速度慢、学习能力难以持续提升等问题。该研究嵌入规则的多智能体强化学习技术,提出基于组合训练的规则与学习结合的方式,分别设计融合规则的多智能体强化学习模型与规则选择模型,通过组合训练将两者有机结合,能够根据当前态势决定使用强化学习决策还是使用规则决策,有效解决在学习中使用哪些规则以及规则使用时机的问题。依托中国电子科技集团发布的多智能体对抗平台,对提出的方法进行实验分析和验证。通过与内置对手对抗,嵌入规则的方法经过约1.4万局训练就收敛到60%的胜率,而没有嵌入规则的算法需要约1.7万局的时候收敛到50%的胜率,结果表明嵌入规则的方法能够有效提升学习的收敛速度和最终效果。

关键词：多智能体强化学习嵌入规则规则选择模型组合训练

Adaptive Hybrid Forward Error Correction Coding Scheme for Video Transmission

在线全文

学校读者我要写书评

暂无评论

ZTE Communications 2024年第2期22卷 85-93页

作者： XIONG Yuhui LIU Zhilong xu Lingmin HUA xinhai WANG Zhaoyang BI Ting JIANG Tao Huazhong University of Science and Technology Wuhan 430074China ZTE Corporation Shenzhen 518057China

This paper proposes an adaptive hybrid forward error correction(AH-FEC)coding scheme for coping with dynamic packet loss events in video and audio ***,the proposed scheme consists of a hybrid Reed-Solomon and low-density parity-check(RS-LDPC)coding system,combined with a Kalman filter-based adaptive *** hybrid RS-LDPC coding accommodates a wide range of code length requirements,employing RS coding for short codes and LDPC coding for medium-long *** delimit the short and medium-length codes by coding performance so that both codes remain in the optimal ***,a Kalman filter-based adaptive algorithm has been developed to handle dynamic alterations in a packet loss *** Kalman filter estimates packet loss rate utilizing observation data and system models,and then we establish the redundancy decision module through receiver *** a result,the lost packets can be perfectly recovered by the receiver based on the redundant *** results show that the proposed method enhances the decoding performance significantly under the same redundancy and channel packet loss.

关键词： video transmission packet loss Reed-Solomon code Kalman filter

同方期刊数据库

在线全文

学校读者我要写书评

暂无评论

基于三维剖分的机载雷达实时探测仿真方法

系统仿真学报 2023年第2期35卷 268-276页

作者：徐颖张帅谢智歌徐新海孙曼晖郭宁军事科学院战争研究院北京100091

无人机的出现及迅猛发展,使得作战仿真中无人机机载雷达的目标探测问题具备较大的研究价值。国内外已有的作战仿真平台中,雷达与目标间探测关系的计算是两两交互式的,计算开销随实体数量呈线性或超线性增长,难以应对大规模实时作战仿真... 详细信息

无人机的出现及迅猛发展,使得作战仿真中无人机机载雷达的目标探测问题具备较大的研究价值。国内外已有的作战仿真平台中,雷达与目标间探测关系的计算是两两交互式的,计算开销随实体数量呈线性或超线性增长,难以应对大规模实时作战仿真场景。基于三维剖分网格的理念,提出一种机载雷达目标探测仿真方法,通过在仿真前制作探测模板,仿真中查询模板及简要计算,可迅速判断探测成败。该方法有效降低了探测的计算量,为大规模实时作战仿真提供了有力支撑。

关键词：作战仿真机载雷达目标探测三维剖分网格

MPtostream:an OpenMP compiler for CPU-GPU heterogeneous parallel systems

在线全文

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2012年第9期55卷 1961-1971页

作者： YANG xueJun,TANG Tao ,WANG GuiBin,JIA Jia & xu xinhai National Laboratory for Parallel and Distributed Processing,National University of Defense Technology,Changsha 410073,China National Laboratory for Parallel and Distributed Processing National University of Defense Technology Changsha China

In light of GPUs’ powerful floating-point operation capacity,heterogeneous parallel systems incorporating general purpose CPUs and GPUs have become a highlight in the research field of high performance computing(HPC).However,due to the complexity of programming on GPUs,porting a large number of existing scientific computing applications to the heterogeneous parallel systems remains a big *** OpenMP programming interface is widely adopted on multi-core CPUs in the field of scientific *** effectively inherit existing OpenMP applications and reduce the transplant cost,we extend OpenMP with a group of compiler directives,which explicitly divide tasks among the CPU and the GPU,and map time-consuming computing fragments to run on the GPU,thus dramatically simplifying the *** have designed and implemented MPtoStream,a compiler of the extended OpenMP for AMD’s stream processing *** experimental results show that programming with the extended directives deviates from programming with OpenMP by less than 11% modification and achieves significant speedup ranging from 3.1 to 17.3 on a heterogeneous system,incorporating an Intel Xeon E5405 CPU and an AMD FireStream 9250 GPU,over the execution on the Xeon CPU alone.

关键词： GPGPU stream OpenMP compiler