Design method, device and storage medium of federated learning mechanism

Abstract

The invention discloses a design method of a federated learning mechanism. The method includes: configuring a preset calculation graph; acquiring an optional optimization goal; configuring a loss function for the preset calculation graph according to the optional optimization goal, and obtaining a calculation graph configured with a loss function; obtaining a priori distribution of data validity Data and participant type prior distribution data; generate random training data according to the data validity prior distribution data and the participant type prior distribution data; train the configuration with loss function calculation according to the random training data graph, to obtain the optimal training parameters; according to the optimal training parameters and the preset calculation graph, a federated learning mechanism is obtained. The invention also discloses a design device of a federated learning mechanism and a computer-readable storage medium. The invention can realize the convenience of improving the calculation of the federated learning mechanism.

Description

Technical field [0001] The present invention relates to the field of mechanism design, and more particularly to a federal learning mechanism design method, a device, and computer readable storage medium. Background technique [0002] Federated Learning is also known as the federal machine learning, is a machine learning framework that can effectively help multiple agencies perform data use and machine learning modeling under the requirements of user privacy protection, data security and government regulations. Federal learning allows multiple data providers to be combined to model, enough to solve the problem of data islandization. At present, federal learning needs to design an incentive mechanism to identify an optimal social selection function, which is used to meet the premise of the requirements of the mechanism on excitation compatibility, individual rationality, budget balance, social welfare, and equity and equity. Under, different economic excitation are provided for dif...

AI Technical Summary

Problems solved by technology

[0003] But at present, in the existing technology, the design process of the federated learning incentive mechanism, that is, the process of determining the optimal social choice function, still relies on the experience and intuition of senior mechanism design experts, which will not only result in inaccurate design results, but also require A certain amount of human and mental work has limitations

Images