Trainieren eines Encodermodells

Details

Veröffentlichungs-Nr.

EP4779522

Anmeldetag

16. Dezember 2025

Veröffentlichung

22. Juli 2026

Rechtsraum

EP

Abstract

There is provided a method of training an encoder model to map inputs representing nominally healthy assets to a cluster of data points. The method comprises providing an input representing a behaviour of an asset at a first point in time to the encoder model undergoing training to map the input to a data point in the feature space, wherein the encoder model is intended to cause data points relating to measured values of input parameters acquired from the same asset at similar time to cluster in feature space, providing the data point to a decoder model, to determine a reconstructed input, determining a difference between the input, and the reconstructed input, and training the encoder model based on the determined difference to map inputs representing nominally healthy assets acquired at similar times to a cluster of data points.

Anmelder

Firma

Rolls-Royce plc

Land

🇬🇧 Großbritannien