[mindcrime]

Offhand, I would have thought that would be limited more to knowledge about audio signal processing and what-not, than the actual machine learning itself. I mean, yeah, you can get REALLY deep in the weeds with neural networks, but you can do an awful lot without doing so. If you ever feel like writing about this, I think a lot of people might find value in a blog post from you about what challenges you ran into and what you found out so far. I know I would.

[fspear]

Yes you are absolutely right, DSP knowledge is also a must. I didn't get that far, I started by identifying what makes people like a certain song and got it down to: Lyrical themes (and overall artist thematic) and melodic and rhythmic patterns. For instance by analyzing data from last.fm I found that a lot of people that listen to Black Metal also seem to have an interest in european/celtic/gaelic folk music, but not all only certain artists with specific lyrical themes and what seemed to me as specific melodic patterns.

Anyway the thing that stumped me was identifying/creating the "features" for rhythmic & melodic patterns, eventually I realized that I was way out of my league and had to drop it.

Later on I found that someone else was already working on this: http://benanne.github.io/2014/08/05/spotify-cnns.html