The following is the repositry containing the python note book of the code which I used for generation of music from ABC format using Character Based Prediction and LSTMs.
The comparative results of Simple Recurrent Networs(SRNs), Long Short Term Memory(LSTMs) and Gated Recurrent Units(GRUs) with different number of epochs have been shown in the Results.doc
Corresponding to each experiment number in the Results.doc is the generated ABC file in GeneratedOutputs. EasyABC software(open source) could be used to hear the MIDI versions of the generated audio.
The experiment has been done by using different layers of unrollability of LSTMs as well as with different stacking of LSTMs.
The code took help from many open sourced projects and research papers.