Cormorant is an experimental pipeline to provide an alternative to standard recommendation systems.
Most recommendation systems, such as Spotify and Youtube, operate on a fairly simple priciple: if person A like Song 1 and Song 2, then if person B like Song 1, they have a chance of liking Song 2.
Cormorant does not do this. Instead, Cormorant will attempt to look at the spectrogram of various songs and determine patterns as to which songs I like, and which ones I do not. Think of this is as an objective differentiaton instead of a subjective or comparative one.
- Viability - Spectrogram test and spider test for youtube.
- Training the Autoencoder
- Setting up data - songs I like, songs I don't. Probably want an interface and some kind of backend - Mongo?