Chapter 09 · Sustainability & AISustainability & AI

Federated learning

Meaning statusEstablishedSource recordDirect document linkedWhy these are different

Definition

A machine-learning approach in which a shared model is trained across many devices or organisations without centralising their raw data.

References

AISTATS 2017 / arXivMcMahan et al. — Communication-Efficient Learning from Decentralized Data

This source provides part of the technical or institutional basis for the definition.

UK ICOPrivacy-enhancing technologies guidance (incl. synthetic data)

This source supports the explanation of how the term is applied, measured or governed in practice.

Overview

What it means

Introduced by Google researchers in 2017, federated learning sends model updates — not data — from participants to a coordinating server, which aggregates them. Privacy is improved by design, though updates can still leak information without additional safeguards.

How it is used

Proposed sustainability uses include training across farms, utilities or factories that cannot pool raw operational data, and mobility or energy models built from devices that keep personal data local.

Why it matters

Data about emissions, yields and consumption is often commercially or legally sensitive. Federated learning offers a route to shared models without shared data — relevant wherever confidentiality blocks collaborative sustainability analytics.