Hey dexed is awesome! I was the coauthor for [2], it for was my 3rd year undergraduate dissertation. A lot of the work went into creating a Python VST host capable of generating huge datasets of random patches.

I'm looking at some of the fancy new ML tricks that have come out since that paper was wrote and have been thinking there would be significant improvements over the LSTM we used.