NOVELTY DETECTION METHOD FOR A NEURAL NETWORK MODEL

NOVELTY DETECTION METHOD FOR A NEURAL NETWORK MODEL

A computer-implemented method (12) for detecting, for a neural network model configured to classify input signals into a plurality of known output classes, whether an input signal (x) belongs to a new output class, the new output class being an output class which does not belong to said plurality of...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: OLMEDA REINO, Daniel, TURNER, Richard, CHUMERIN, Nikolay, AL JUNDI, Rahaf
Format: Patent
Sprache:eng ; fre ; ger
Schlagworte:
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
PHYSICS
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:A computer-implemented method (12) for detecting, for a neural network model configured to classify input signals into a plurality of known output classes, whether an input signal (x) belongs to a new output class, the new output class being an output class which does not belong to said plurality of known output classes, the method comprising:- computing (22) a feature vector (Φ(x)) of the input signal with the neural network model;- calculating (24, 30) a score (B) based on respective similarities between said feature vector (Φ(x)) and respective prototypes of all the known output classes of the neural network model;- determining (32) that the input signal belongs to a new output class if the score (B) meets a predetermined criterion.