Aug 31, 2018 · 1 min read
Sorry Daniel, I am not yet ready to release my code for various reasons. If you want, you can send me an mp3/wav file, and I will encode/decode it for you with 12 frequencies per octave, and see if you can tell the difference ;)
I’ve been reading up on LSTMs, but never used them in practice. It seems like there have been some recent advances related to “attention”, which are supposed to work better than LSTMs, getting rid of the recurrence completely (see “Attention is All You Need”, https://arxiv.org/abs/1706.03762). People seem to usually apply this to text or pictures, there is really almost no activity for music. I think I would like to understand the models mode before trying to apply them to music.
