基于软件定义网络(software defined network,简称SDN)的数据中心流量工程,能够通过对全局视图的网络管控,动态选择路由路径,规避拥塞发生的风险.但是在制定路由策略时,经常会对数据流进行迁移,尤其是针对大流的迁移容易造成数据流丢包...基于软件定义网络(software defined network,简称SDN)的数据中心流量工程,能够通过对全局视图的网络管控,动态选择路由路径,规避拥塞发生的风险.但是在制定路由策略时,经常会对数据流进行迁移,尤其是针对大流的迁移容易造成数据流丢包以及接收端数据包乱序的问题.提出了基于时隙的流片装箱算法(flowlet-binned algorithm based on timeslot,简称FLAT),通过集中控制的方式获取链路状态信息并计算出合理的数据流传输时隙值,能够避免在数据流迁移过程中的丢包以及接收端数据包乱序问题;同时,在充分利用数据中心冗余链路的前提下,实现高效和细粒度的流量均衡.通过在Mininet仿真平台中部署并与ECMP以及GFF路由机制相比较,在链路高负载情况下,丢包率分别下降了90%和80%,而吞吐量分别能够提升44%和11%,实验结果展示了FLAT的优越性能.展开更多
本文提出一种基于UD(upper-diagonal)分解与偏差补偿结合的辨识方法,用于变量带误差(errors-in-variables,EIV)模型辨识.考虑单输入单输出(single input and single output,SISO)线性动态系统,当输入和输出含有零均值、方差未知的高斯...本文提出一种基于UD(upper-diagonal)分解与偏差补偿结合的辨识方法,用于变量带误差(errors-in-variables,EIV)模型辨识.考虑单输入单输出(single input and single output,SISO)线性动态系统,当输入和输出含有零均值、方差未知的高斯测量白噪声时,该类系统的模型参数估计是一种典型的EIV模型辨识问题.为了获得这种EIV模型参数的无偏估计,本文先推导出最小二乘模型参数估计偏差量与输入输出噪声方差以及最小二乘损失函数与输入输出噪声方差的关系,然后采用UD分解方法递推获得模型参数估计值,再利用输入输出噪声方差估计值补偿模型参数估计偏差,以此获得模型参数的无偏估计.本文还讨论了算法实现过程中遇到的一些问题及修补方法,并通过仿真例验证了所提辨识方法的有效性.展开更多
The development of machine learning in complex system is hindered by two problems nowadays.The first problem is the inefficiency of exploration in state and action space,which leads to the data-hungry of some state-of...The development of machine learning in complex system is hindered by two problems nowadays.The first problem is the inefficiency of exploration in state and action space,which leads to the data-hungry of some state-of-art data-driven algorithm.The second problem is the lack of a general theory which can be used to analyze and implement a complex learning system.In this paper,we proposed a general methods that can address both two issues.We combine the concepts of descriptive learning,predictive learning,and prescriptive learning into a uniform framework,so as to build a parallel system allowing learning system improved by self-boosting.Formulating a new perspective of data,knowledge and action,we provide a new methodology called parallel learning to design machine learning system for real-world problems.展开更多
Hybrid precoding can reduce the number of required radio frequency(RF)chains in millimeter-Wave(mmWave) massive MIMO systems. However, existing hybrid precoding based on singular value decomposition(SVD) requires the ...Hybrid precoding can reduce the number of required radio frequency(RF)chains in millimeter-Wave(mmWave) massive MIMO systems. However, existing hybrid precoding based on singular value decomposition(SVD) requires the complicated bit allocation to match the different signal-to-noise-ratios(SNRs) of different sub-channels. In this paper,we propose a geometric mean decomposition(GMD)-based hybrid precoding to avoid the complicated bit allocation. Specifically,we seek a pair of analog and digital precoders sufficiently close to the unconstrained fully digital GMD precoder. To achieve this, we fix the analog precoder to design the digital precoder, and vice versa. The analog precoder is designed based on the orthogonal matching pursuit(OMP) algorithm, while GMD is used to obtain the digital precoder. Simulations show that the proposed GMD-based hybrid precoding achieves better performance than the conventional SVD-based hybrid precoding with only a slight increase in complexity.展开更多
Recent progress of research for graphene applications in electronic and optoelectronic devices is reviewed, and recent developments in circuits based on graphene devices are summarized. The bandgap-mobility tradeoff i...Recent progress of research for graphene applications in electronic and optoelectronic devices is reviewed, and recent developments in circuits based on graphene devices are summarized. The bandgap-mobility tradeoff inevitably constrains the application of graphene for the conventional field-effect transistor (FET) devices in digital applications. However, this shortcoming has not dampened the enthusiasm of the research community toward graphene electronics. Aside from high mobility, graphene offers numerous other amazing electrical, optical, thermal, and mechanical properties that continually motivate innovations.展开更多
In order to improve the scalability and reliability of Software Defined Networking(SDN),many studies use multiple controllers to constitute logically centralized control plane to provide load balancing and fail over.I...In order to improve the scalability and reliability of Software Defined Networking(SDN),many studies use multiple controllers to constitute logically centralized control plane to provide load balancing and fail over.In this paper,we develop a flexible dormant multi-controller model based on the centralized multi-controller architecture.The dormant multi-controller model allows part of controllers to enter the dormant state under light traffic condition for saving system cost.Meanwhile,through queueing analysis,various performance measures of the system can be obtained.Moreover,we analyze the real traffic of China Education Network and use the results as the parameters of computer simulation and verify the effects of parameters on the system characteristics.Finally,a total expected cost function is established,and genetic algorithm is employed to find the optimal values of various parameters to minimize system cost for the deployment decision making.展开更多
An improved fruit fly optimization algorithm( iFOA) is proposed for solving the lot-streaming flow-shop scheduling problem( LSFSP) with equal-size sub-lots. In the proposed iFOA,a solution is encoded as two vectors to...An improved fruit fly optimization algorithm( iFOA) is proposed for solving the lot-streaming flow-shop scheduling problem( LSFSP) with equal-size sub-lots. In the proposed iFOA,a solution is encoded as two vectors to determine the splitting of jobs and the sequence of the sub-lots simultaneously. Based on the encoding scheme,three kinds of neighborhoods are developed for generating new solutions. To well balance the exploitation and exploration,two main search procedures are designed within the evolutionary search framework of the iFOA,including the neighborhood-based search( smell-vision-based search) and the global cooperation-based search. Finally,numerical testing results are provided,and the comparisons demonstrate the effectiveness of the proposed iFOA for solving the LSFSP.展开更多
文摘基于软件定义网络(software defined network,简称SDN)的数据中心流量工程,能够通过对全局视图的网络管控,动态选择路由路径,规避拥塞发生的风险.但是在制定路由策略时,经常会对数据流进行迁移,尤其是针对大流的迁移容易造成数据流丢包以及接收端数据包乱序的问题.提出了基于时隙的流片装箱算法(flowlet-binned algorithm based on timeslot,简称FLAT),通过集中控制的方式获取链路状态信息并计算出合理的数据流传输时隙值,能够避免在数据流迁移过程中的丢包以及接收端数据包乱序问题;同时,在充分利用数据中心冗余链路的前提下,实现高效和细粒度的流量均衡.通过在Mininet仿真平台中部署并与ECMP以及GFF路由机制相比较,在链路高负载情况下,丢包率分别下降了90%和80%,而吞吐量分别能够提升44%和11%,实验结果展示了FLAT的优越性能.
文摘本文提出一种基于UD(upper-diagonal)分解与偏差补偿结合的辨识方法,用于变量带误差(errors-in-variables,EIV)模型辨识.考虑单输入单输出(single input and single output,SISO)线性动态系统,当输入和输出含有零均值、方差未知的高斯测量白噪声时,该类系统的模型参数估计是一种典型的EIV模型辨识问题.为了获得这种EIV模型参数的无偏估计,本文先推导出最小二乘模型参数估计偏差量与输入输出噪声方差以及最小二乘损失函数与输入输出噪声方差的关系,然后采用UD分解方法递推获得模型参数估计值,再利用输入输出噪声方差估计值补偿模型参数估计偏差,以此获得模型参数的无偏估计.本文还讨论了算法实现过程中遇到的一些问题及修补方法,并通过仿真例验证了所提辨识方法的有效性.
基金Supported by National Basic Research Program of China (973 Program) (2010CB731800) and National Natural Science Foundation of China (60974059, 60736026, 61021063)
基金Supported by National High Technology Research and Development Program of China (863 Program) (2006AA04Z42g), National Natural Science Foundation of China (60574085, 60736026, 60721003), and German Research Foundation (DI 773/10)
基金Supported by National Natural Science Foundation of China(61071131,61271388) Natural Science Foundation of Beijing(4122040)+1 种基金 Research Project of Tsinghua University(2012Z01011) Doctoral Fund of Ministry of Education of China(20120002110036)
基金supported in part by the National Natural Science Foundation of China(91520301)
文摘The development of machine learning in complex system is hindered by two problems nowadays.The first problem is the inefficiency of exploration in state and action space,which leads to the data-hungry of some state-of-art data-driven algorithm.The second problem is the lack of a general theory which can be used to analyze and implement a complex learning system.In this paper,we proposed a general methods that can address both two issues.We combine the concepts of descriptive learning,predictive learning,and prescriptive learning into a uniform framework,so as to build a parallel system allowing learning system improved by self-boosting.Formulating a new perspective of data,knowledge and action,we provide a new methodology called parallel learning to design machine learning system for real-world problems.
基金supported by the National Natural Science Foundation of China for Outstanding Young Scholars (Grant No. 61722109)the National Natural Science Foundation of China (Grant No. 61571270)the Royal Academy of Engineering through the UK–China Industry Academia Partnership Programme Scheme (Grant No. UK-CIAPP\49)
文摘Hybrid precoding can reduce the number of required radio frequency(RF)chains in millimeter-Wave(mmWave) massive MIMO systems. However, existing hybrid precoding based on singular value decomposition(SVD) requires the complicated bit allocation to match the different signal-to-noise-ratios(SNRs) of different sub-channels. In this paper,we propose a geometric mean decomposition(GMD)-based hybrid precoding to avoid the complicated bit allocation. Specifically,we seek a pair of analog and digital precoders sufficiently close to the unconstrained fully digital GMD precoder. To achieve this, we fix the analog precoder to design the digital precoder, and vice versa. The analog precoder is designed based on the orthogonal matching pursuit(OMP) algorithm, while GMD is used to obtain the digital precoder. Simulations show that the proposed GMD-based hybrid precoding achieves better performance than the conventional SVD-based hybrid precoding with only a slight increase in complexity.
文摘Recent progress of research for graphene applications in electronic and optoelectronic devices is reviewed, and recent developments in circuits based on graphene devices are summarized. The bandgap-mobility tradeoff inevitably constrains the application of graphene for the conventional field-effect transistor (FET) devices in digital applications. However, this shortcoming has not dampened the enthusiasm of the research community toward graphene electronics. Aside from high mobility, graphene offers numerous other amazing electrical, optical, thermal, and mechanical properties that continually motivate innovations.
基金the National High-tech R&D Program ("863" Program) of China,the National Science Foundation of China,National Science & Technology Pillar Program of China,the National Science Foundation of China,the Post-Doctoral Funding of China,Tsinghua-Huawei joint research project
文摘In order to improve the scalability and reliability of Software Defined Networking(SDN),many studies use multiple controllers to constitute logically centralized control plane to provide load balancing and fail over.In this paper,we develop a flexible dormant multi-controller model based on the centralized multi-controller architecture.The dormant multi-controller model allows part of controllers to enter the dormant state under light traffic condition for saving system cost.Meanwhile,through queueing analysis,various performance measures of the system can be obtained.Moreover,we analyze the real traffic of China Education Network and use the results as the parameters of computer simulation and verify the effects of parameters on the system characteristics.Finally,a total expected cost function is established,and genetic algorithm is employed to find the optimal values of various parameters to minimize system cost for the deployment decision making.
基金National Key Basic Research and Development Program of China(No.2013CB329503)National Natural Science Foundation of China(No.61174189)the Doctoral Program Foundation of Institutions of Higher Education of China(No.20130002110057)
文摘An improved fruit fly optimization algorithm( iFOA) is proposed for solving the lot-streaming flow-shop scheduling problem( LSFSP) with equal-size sub-lots. In the proposed iFOA,a solution is encoded as two vectors to determine the splitting of jobs and the sequence of the sub-lots simultaneously. Based on the encoding scheme,three kinds of neighborhoods are developed for generating new solutions. To well balance the exploitation and exploration,two main search procedures are designed within the evolutionary search framework of the iFOA,including the neighborhood-based search( smell-vision-based search) and the global cooperation-based search. Finally,numerical testing results are provided,and the comparisons demonstrate the effectiveness of the proposed iFOA for solving the LSFSP.