分布式多自主体优化问题中的概率量化影响研究

考虑一个由多个自主体构成的网络,网络中每个自主体拥有一个只有自己知晓的局部目标函数,研究优化所有局部目标函数和的问题.基于一致性算法基本思想并结合次梯度方法解决了固定的网络拓扑结构且自主体之间交换的信息是经过概率量化的分布式多自主体优化问题.得到一个与量化精度和网络连接度有关的关于收敛速率的上界.当步长固定时,该上界可保证网络中的每个自主体的状态值收敛到最优解附近.

Effects of Probabilistic Quantization on Distributed Multi-agent Optimization Problem

This paper studies the problem of optimizing the sum of the individual local convex objective functions which are known to each agent accordingly.Based on the average consensus algorithm,this paper defines a subgradient step to solve this problem when the underlying network topology is fixed and the data transmitted through agents are probabilistically quantized.This paper derives an upper bound which is related to the quantization resolution and the underlying network connectivity on the convergence rate o...