Federal learning system

Abstract

The invention discloses a federation learning system, and the system comprises a federation construction unit, a data mart unit, a model training unit, a model release unit and a report statistics unit, wherein the federation construction unit comprises a federation member module, a platform building module and a federation cooperation module; the data market unit comprises a data shelving module,a data searching module and a subscription management module, the model training unit comprises a visual modeling module, a training task module and a training tracking module, and the model publishing unit comprises a publishing module, a prediction task module and a prediction tracking module; the report statistics unit is used for performing model training frequency statistics, ID intersectionstatistics, model prediction frequency statistics and data use condition statistics on each party of alliance members.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a federated learning system. Background technique [0002] Modern machine learning systems can be vulnerable to a variety of failures, including non-malicious failures such as bugs in preprocessing pipelines, noisy training labels, unreliable clients, and explicit attacks against training and deployment pipelines. The distributed nature, architecture design, and data constraints of federated learning open up new failure modes and attack surfaces. In federated learning, there are data privacy protection problems and risks of data leakage. The requirements for configuring the algorithm process are relatively high, which limits the application of federated learning. Contents of the invention [0003] In order to solve the above problems, the present invention provides a federated learning system. [0004] The present invention adopts following technical scheme: [0005...

AI Technical Summary

Problems solved by technology

The distributed nature, architecture design, and data constraints of federated learning open up new failure modes and attack surfaces. In federated learning, there are data privacy protection problems and risks of data leakage. The requirements for configuring the algorithm process are relatively high, which limits the application of federated learning

Images