Efficient asynchronous federated learning method for reducing communication times

Abstract

The invention relates to an efficient asynchronous federated learning method for reducing communication times. The method comprises the following steps: firstly, designing a hyper-parameter r which adaptively changes along with version obsolescence, reducing errors brought to asynchronous federal learning by the version obsolescence, and guiding model convergence; and to solve the large federated learning communication traffic, increasing the learning rate and the decreasing the local round number in the early stage, and then gradually reducing the learning rate to increase the local round number, so that the performance of the model can be basically unchanged under the condition of effectively reducing the total communication round number of model training, and the system can better carry out asynchronous federated learning.

Description

Technical field: [0001] The invention relates to an efficient asynchronous federated learning method for reducing communication times, and the method has good application in the federated learning field. Background technique: [0002] In the field of asynchronous federated learning, the version staleness of the local model will cause certain errors or even errors for the update of the global model. Controlling the version staleness can reduce the error caused by the version staleness and guide the model to converge. At the same time, in the process of federated learning, the large amount of model parameters leads to huge communication traffic during remote device training, which has always been a problem that plagues federated learning. Generally, people will use methods such as model distillation to reduce the amount of model parameters and reduce the amount of information in each communication. [0003] How to control the traffic in the process of federated learning and c...

AI Technical Summary

Problems solved by technology

At the same time, in the process of federated learning, the large amount of model parameters leads to huge communication traffic during remote device training, which has always been a problem that plagues federated learning.

Images