I love music. I also love computers. Why not combine them in some way to make both even better?

- Everybody Ever, Thinking They're The First

The field of computer-generated music is one of my absolute favorite topics. For anyone who has taken a music theory class, it seems such an obvious conclusion: if there are rules governing how music is structured, can't we simply generate music at will? For anyone who has delved into machine learning, the idea becomes even more intriguing- we don't have to program the rules, we can learn them! Unfortunately, it's not quite so easy.

The Music ex Machina project (here on GitHub) represents my best attempts at computationally-generated music. They range from algorithmic composition (one of my most successful, but most shallow attempts) to a full-on deep learning approach, modeled in part on the wonderful work of Daniel Johnson.