High-energy-efficiency computing communication joint optimization method for edge federated learning

Abstract

The invention discloses a high-energy-efficiency computing communication joint optimization method for edge federated learning. The method comprises the steps that an edge server receives communication / calculation element energy consumption information, sent by each edge equipment, in the edge equipment, and the edge server receives the element energy consumption information, optimizes an energy consumption control scheme according to the element energy consumption information, and issues the optimized energy consumption control scheme to each edge device, so that the edge device configures related communication and calculates the working parameters of the energy consumption element according to the energy consumption control scheme, and then the edge server completes federated learning under the cooperation of the edge equipment. According to the invention, the energy consumption distribution and the working condition can be adaptively adjusted according to the hardware condition andthe energy consumption condition of the edge equipment, and the reasonable resource redistribution setting of the energy consumption element is subjected to energy efficiency optimization, so that thepurposes of reducing the energy consumption and improving the utilization efficiency of the energy are achieved.

Description

technical field [0001] This application relates to the field of communication optimization, in particular to an edge federated learning-oriented joint optimization method for computing and communication with high energy efficiency. Background technique [0002] In 2016, Alphago defeated the human chess player Li Shidol, which set off a wave of artificial intelligence all over the world. However, the training of artificial intelligence models has strict requirements on computing power, so that most of the traditional artificial intelligence algorithm operations and model training can only be concentrated. in the cloud computing center. In order to solve this problem, Google has proposed a machine learning solution with decentralized training data, namely "Federated Learning", which aims to learn a high The quality-centralized machine learning model solves the problem of data islands. The overall process of "federated learning" is as follows: the edge server first sends the ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of this application is to provide an energy-efficient computing and communication joint optimization method for edge federated learning to solve the problem of computing and communication energy consumption behind "federal learning" under the premise of satisfying time constraints and accuracy constraints

Images