LEARNING TO RANK MUSIC TRACKS USING TRIPLET LOSS - Equipe Signal, Statistique et Apprentissage Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

LEARNING TO RANK MUSIC TRACKS USING TRIPLET LOSS

Résumé

Most music streaming services rely on automatic recommendation algorithms to exploit their large music catalogs. These algorithms aim at retrieving a ranked list of music tracks based on their similarity with a target music track. In this work, we propose a method for direct recommendation based on the audio content without explicitly tagging the music tracks. To that aim, we propose several strategies to perform triplet mining from ranked lists. We train a Convolutional Neural Network to learn the similarity via triplet loss. These different strategies are compared and validated on a large-scale experiment against an auto-tagging based approach. The results obtained highlight the efficiency of our system, especially when associated with an Auto-pooling layer.
Fichier principal
Vignette du fichier
camera_ready.pdf (333.54 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02477242 , version 1 (13-02-2020)

Identifiants

  • HAL Id : hal-02477242 , version 1

Citer

Laure Prétet, Gael Richard, Geoffroy Peeters. LEARNING TO RANK MUSIC TRACKS USING TRIPLET LOSS. ICASSP, May 2020, Barcelona, Spain. ⟨hal-02477242⟩
252 Consultations
842 Téléchargements

Partager

Gmail Facebook X LinkedIn More