Dual-track Music Generation using Deep Learning

TL;DR

This paper introduces a dual-track deep learning architecture for classical piano music generation that models the inter-dependency between left and right hands, outperforming previous methods and including novel training policies.

Contribution

The paper presents a new dual-track neural network architecture for piano music generation, with innovative training policies and comprehensive evaluation against existing models and real music.

Findings

Proposed model outperforms tested methods

Special training policies improve performance

Model comparison with MuseGAN and real music

Abstract

Music generation is always interesting in a sense that there is no formalized recipe. In this work, we propose a novel dual-track architecture for generating classical piano music, which is able to model the inter-dependency of left-hand and right-hand piano music. Particularly, we experimented with a lot of different models of neural network as well as different representations of music, and the results show that our proposed model outperforms all other tested methods. Besides, we deployed some special policies for model training and generation, which contributed to the model performance remarkably. Finally, under two evaluation methods, we compared our models with the MuseGAN project and true music.