Federal learning model training method, client, server and storage medium

Abstract

The invention relates to the technical field of artificial intelligence algorithms, and particularly provides a federated learning model training method based on knowledge distillation, which comprises the following steps: receiving control parameters for model training from a server; training the initial first neural network model according to the control parameter and the local data sample to obtain a first model parameter; sending the first model parameter to the server; receiving a second model parameter of a second neural network model from the server; and utilizing a knowledge distillation method to enable the first neural network model to learn knowledge of the second neural network model, and training to obtain an updated first neural network model. According to the method, the existing data island problem is effectively solved by constructing the federated learning system, and meanwhile, the knowledge distillation module is added in the federated learning framework system, so that the algorithm model can be trained and optimized on the basis of knowledge of all training data at the same time; and the training effect of the federal learning framework system is further improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a federated learning model training method based on knowledge distillation, a client, a server, and a computer-readable storage medium. Background technique [0002] At present, if multiple units cooperate to use artificial intelligence algorithms to implement in a certain business scenario, they will encounter some problems. For example, due to data security and data privacy requirements, the data of each unit cannot be effectively circulated among units. use, resulting in the problem of data islands. The traditional algorithm training framework emphasizes the diversity and integrity of data, which further amplifies the impact of data islands on algorithm capabilities. Therefore, the traditional algorithm training framework and the data island problem will make the artificial intelligence algorithm capability fall into a bottleneck, and further limit the...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problem that the existing solutions cannot improve the algorithm ability due to data islands and limit the application of the algorithm, on the one hand, a federated learning model training method based on knowledge distillation includes: receiving The control parameters used for model training acquired by the server; training the initial first neural network model according to the control parameters and local data samples to obtain first model parameters; sending the first model parameters to the server; Receiving the second model parameters of the second neural network model obtained by the server; using the knowledge distillation method to enable the first neural network model to learn the knowledge of the second neural network model, and train to obtain an updated first neural network model , wherein the first neural network model is a student network model

Images