Variance reduction technology-based distributed projection method considering communication delay

Abstract

The invention discloses a variance reduction technology-based distributed projection method considering communication time delay. The method comprises the following steps of 1, proposing an original optimization problem model (1) for a multi-intelligent system with local set constraints and local equation constraints at the same time; 2, equivalently converting the original optimization problem model (1) obtained in the step 1 into a convex optimization problem model (2) convenient for distribution processing; step 3, proposing a variance reduction technology-based distributed projection algorithm (3), solving a constrained convex optimization problem model (2), i.e., estimating a local full gradient in an unbiased manner by adopting a local random average gradient so as to reduce a heavycalculation burden caused by calculating full gradients of all local target functions in each iteration; and 4, performing convergence analysis. According to the invention, the calculation cost of allagents in the network can be greatly reduced, so that the communication and calculation pressure of the whole multi-agent system is reduced, and the practicability is relatively high.

Description

technical field [0001] The present invention relates to the technical field of intelligent communication, in particular to a distributed projection method based on variance reduction technology considering communication time delay. Background technique [0002] In recent years, with the vigorous development of high technology, especially the emergence of emerging fields such as cloud computing and big data. Distributed optimization theory and applications have received more and more attention, and gradually penetrated into all aspects of scientific research, engineering applications, and social life. Distributed optimization is effectively optimized through the cooperation and coordination among multiple agents. It can be used to solve large-scale and complex optimization problems that are difficult for many centralized algorithms. However, when the existing distributed optimization algorithms face large-scale convex optimization problems with relatively complex local const...

AI Technical Summary

Problems solved by technology

However, when the existing distributed optimization algorithms face large-scale convex optimization problems with relatively complex local constraints, the amount of gradient calculation is large, and the calculation burden of the agents in the network is heavy, which leads to the calculation and communication of multi-agent systems. Inefficiency and other deficiencies, so it cannot meet people's requirements

Images