Federal machine learning method and server

Abstract

The invention provides a method of federated machine learning using at least one processor, the method comprising: transmitting a current global machine learning model to each of a plurality of data sources; receiving a plurality of training updates from the plurality of data sources, respectively, each of the plurality of training updates being generated by the respective data source in response to the received global machine learning model; and update the current global machine learning model based on the received plurality of training updates and a plurality of data quality parameters associated with the plurality of data sources, respectively, to generate an updated global machine learning model. A corresponding server for federated machine learning is also provided.

Description

technical field [0001] The present invention generally relates to a method for federated machine learning and a server thereof. Background technique [0002] Supervised deep learning algorithms provide state-of-the-art performance for various classification tasks, such as image classification tasks. The traditional approach to these tasks may involve three steps: (a) centralizing large data repositories, (b) obtaining ground-truth annotations for these data, and (c) using ground-truth annotations to train a convolutional neural network (CNN) for classification , however, this framework poses significant practical challenges. [0003] In particular, data privacy as well as security concerns make it difficult to create large central data repositories for training. Recent work has developed decentralized federated learning approaches to train deep learning models across multiple data sources without sharing sensitive information. These existing federated learning approaches ...

AI Technical Summary

Problems solved by technology

These existing federated learning approaches have proven successful, but may still suffer from inaccuracy and / or unreliability, depending on the data source on which they are trained

Images