Hacker News Reader

I've been listening to the Stanford NLP deep learning lectures, and while I know hardly anything about it compared to most, "deep learning" is screaming at me. If you could generate a "corpus" of musical progressions based off of real music, you could very easily sample from a model to generate progressions that are likely to "co-occur."

This was posted three months ago on HN, but one person used it to generate music. http://karpathy.github.io/2015/05/21/rnn-effectiveness/

The folk music generation is located at https://soundcloud.com/seaandsailor/sets/char-rnn-composes-i.... I could imagine you could do something similar.

Do you need to generate an unlimited number of them ? Could you just listen to the radio for a few hours and write down some patterns?

I know next to nothing about music theory.

In the example he listed each note goes either up or down one note to the next chord.