Learning to rank music tracks using triplet loss
This work addresses music recommendation for streaming services by proposing a direct audio-based method, but it appears incremental as it builds on existing triplet loss and CNN techniques.
The authors tackled the problem of ranking music tracks for recommendation by training a Convolutional Neural Network with triplet loss using strategies for triplet mining from ranked lists, achieving results that highlight the efficiency of their system, especially with an Auto-pooling layer.
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.