A neural network approach to generating polyphonic musical compositions
This is my term project for Fall 2018 UMass CS682 - Neural Networks. The goal is to use unsupervised feature learning to create a method of musical style transfer between composers.
I created a convolutional neural network for composer classification, and used a gram matrix method to perform style transfer between two pieces from different composers.
To view the scores, which are written in MusicXML, you need a reader such as the free MuseScore.