This is the result of a project I worked on for CS224D with Aran Nayebi. The idea is to design a neural network that can generate music using your music library as training data. While I believe we could probably improve upon this model significantly, it serves as a good proof of concept for showing that it is indeed possible. We use Long Short-Term Memory (LSTM) networks to model hidden recurrences over time. For this demo, we used a network that was trained on a variety of Madeon songs.
UPDATE:
Due to popular demand, we've released the source code on GitHub. Check it out here: https://github.com/MattVitelli/GRUV
Happy training! :D