Dezentrales System zum Föderierten Lernen

Details

Veröffentlichungs-Nr.

EP3786872

Aktenzeichen

EP20187465

Anmeldetag

23. Juli 2020

Veröffentlichung

30. Juli 2025

Erteilung

30. Juli 2025

Rechtsraum

EP

IPC

G06Q20/38G06N3/08H04L9/00

Abstract

A participant node of a distributed ledger network may identify a distributed federated learning (DFL) smart contract stored on a blockchain. The DFL smart contract may include an aggregation sequence. The aggregation sequence may include an ordered sequence of participant node identifiers. The participant node may generate a trained model by training a global model with training data. The participant node may detect, on the blockchain, a first transition token indicative of a first model previously aggregated by another participant node. The participant node may receive the first model. The participant node may aggregate the first model with the trained model to generate a second model. The participant node may store, on the blockchain, a second transition token indicative of the second model. A successor node identified in the aggregation sequence may further aggregate the second model with an additional model in response to detection of the second transition token.

Anmelder

Firma

Accenture Global Solutions Limited

Land

🇮🇪 Irland

Fachgebiete