Contrastive Self-supervised learning for music
Computational Music Analysis and Recommendation
Contrastive learning is very promising for music because it doesn't require labels. Or doesn't it? Still, we need categories that would define the contrast - albums, artists, genres, instruments. This thesis is an exploratory work to find which contrastive learning scheme works best for taste profiles prediction.
Graduation Theses defence year
Spoken language (s)
Requirements for candidates
Application of contact