Federal learning system and method based on noise distillation

Abstract

The invention provides a federated learning system and method based on noise distillation. The federated learning system based on noise distillation comprises the steps that at least two clients determine a local business sample and an initial business model; updating the model into an intermediate business model according to a preset constraint strategy; inputting the local service sample into the initial service model and the intermediate service model respectively to obtain initial prediction information and intermediate prediction information; performing parameter adjustment on the intermediate service model based on a local sample label corresponding to the local service sample, the initial prediction information and the intermediate prediction information until a target service model meeting a condition is obtained; parameters corresponding to the model are sent to a server; the server receives the model parameters; constructing an initial global business model according to the model parameters, and constructing a noise business sample corresponding to each model parameter; and training the initial global service model by using the noise service sample until a target global service model meeting a service training stop condition is obtained.

Description

technical field [0001] This specification relates to the field of machine learning technology, in particular to a federated learning system and method based on noise distillation. Background technique [0002] With the development of Internet technology, federated learning has become the main means of solving data silos in the existing technology, while ensuring the security of distributed machine privacy. Most of the common federation algorithms are implemented based on the FedAvg paradigm proposed by the enterprise. However, as the complexity of data and the amount of model parameters increase, the data traffic and the number of iteration rounds when the model is trained with this type of algorithm will be significantly increased, and the algorithm efficiency in federated learning has become a bottleneck. Therefore, an effective solution is urgently needed to solve the above problems. SUMMARY OF THE INVENTION [0003] In view of this, the embodiments of this specificat...

AI Technical Summary

Problems solved by technology

However, as the complexity of the data and the number of model parameters increase, the data communication volume and the number of iterations when the

Images