Chord Progression and Music Estimation Using LSTM Networks

Recurrent Neural Networks and Long-Short Term Memory Units have been very successful in analyzing and handling time-series data with long or short term dependencies. Music is a great domain for these methods to be practiced and applied, as it is comprised of vast amount of time-series data with all kinds of dependencies. This project aims to investigate such methods for tasks like chord progression estimation and autonomus musical composition. Such methods will be practiced on midi formatted musical files and musical data with chroma representations.