Federated learning API

Note

This document has been machine translated.

This section describes the Federated learning API.

Federated Learning

Federated learning is a method of improving model accuracy while maintaining data security by aggregating models trained on multiple clients at the server.

xData Edge provides an API for transferring models stored in the model store between xData Edges in order to achieve federated learning by coordinating multiple xData Edges.

Upstream Information

The xData Edge Federated learning API allows specifying, for each model store, the endpoint URL and model store name of the upper node to which the model will be fed back. This is called the upstream information of the model store.

When a model is transferred (fed back) to a high-level node, it is transferred according to the pre-defined upstream information. When a model is transferred (transferred) to an aggregated model at a high-level node, it is also transferred according to the same pre-defined upstream information.

Federated learning API List

xData Edge users can use the Federated learning API to perform the following operations.

API Methods	Description
fl.feedback	model Feedback a model in a store to an upstream model store.
fl.request_transfer	Retrieve the latest aggregated model from the upstream model store and store it in the model store.
fl.create_upstream_info	register the model store of the upstream node to which the model is fed back in the model store.
fl.update_upstream_info	Update upstream information registered in the model store.
fl.delete_upstream_info	Deletes upstream information from the model store for the upstream node to which the model is fed back.
fl.list_upstream_info	List upstream information in the model store.
fl.get_upstream_info	Retrieve upstream information of a model store.