摘要
针对智能反射面(IRS)辅助的边缘智能系统中模型参数汇聚的问题,提出一种基于数据重要性感知的资源分配算法.利用凸优化和分支定界等方法交替优化用户的发射功率、传输次数和智能反射面的相移矩阵.仿真结果表明,所提算法能够基于本地数据的重要性差异有效汇聚分布式智能体的模型参数,并最大化加权和速率.
In order to solve the problem of model aggregation in intelligent reflecting surface(IRS)aided edge intelligent system,a data-importance-aware resource allocation algorithm is proposed by using convex optimization and branch-and-bound methods to alternately design the user’s uplink power,transmission time,and the phase shifts of IRS.Simulation results show that the proposed algorithm can effectively aggregate the model parameters of the distributed agents based on the importance difference of local data,and can maximize the uplink weighted sum rate.
作者
田辉
倪万里
王雯
郑景桁
贺硕
TIAN Hui;NI Wan-li;WANG Wen;ZHENG Jing-heng;HE Shuo(State Key Laboratory of Networking and Switching Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China;School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China)
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
2020年第6期51-58,共8页
Journal of Beijing University of Posts and Telecommunications
基金
国家重点研发计划项目(2019YFC1511400)
关键词
智能反射面
模型汇聚
重要性感知
资源分配
intelligent reflecting surface
model aggregation
importance aware
resource allocation
