Federal learning method and system for data non-independent identically distributed scene

Abstract

The invention discloses a federal learning method and system for a data non-independent identically distributed scene. The system comprises a plurality of clients and a central server. The central server is used for dividing a target data set into a plurality of sub-data sets in a non-independent distribution mode, so that each sub-data set contains all types of data, and distributing the sub-data sets to the client; the client is used for guiding the sub-data set to train a current local model according to a current local anchor point based on the received sub-data set, updating parameters of the local anchor point and the local model, and uploading model data to the central server according to an agreed communication mode; the central server is further used for aggregating according to the received model data to obtain aggregated data, and downloading the aggregated data to the client according to an agreed communication mode to serve as the basis of the next round of federated learning, the method improves the practicability of the federated learning system in a specific scene on the basis of guaranteeing the safety of user data, and improves the user experience. Meanwhile, the problems of communication efficiency and statistics heterogeneity of a federated learning system are solved.

Description

technical field [0001] The invention belongs to the field of artificial intelligence and information security, and in particular relates to a federated learning method and system for data non-IID scenarios. Background technique [0002] With the continuous in-depth application of new technologies such as big data, artificial intelligence, and cloud computing in various industries, global data is characterized by explosive growth and massive aggregation, and the value of data is becoming more and more prominent. As a production factor, data is faced with two key problems: confirmation of rights and privacy protection. Data is essentially information, not exclusive or exclusive, and can be possessed by most people at the same time. In the era of digital economy, the marginal cost of dissemination of personal-related information is almost zero, and it can quickly spread to the whole world. This low cost makes data protection face special difficulties. At present, companies an...

AI Technical Summary

Problems solved by technology

However, these personalization methods often introduce additional computation and communication overhead that may be unnecessary, and the final results are not satisfactory.

Images