Folk music style modelling by recurrent neural networks with long short term memory units

摘要

We demonstrate two generative models created by traininga recurrent neural network (RNN) with three hiddenlayers of long short-term memory (LSTM) units. This extendspast work in numerous directions, including trainingdeeper models with nearly 24,000 high-level transcriptionsof folk tunes. We discuss our on-going work.