Improving Polyphonic Music Models with Feature-Rich Encoding

TL;DR

This paper demonstrates that enriching sequence representations with salient features and multi-task learning significantly improves polyphonic music modeling performance, achieving state-of-the-art results on the JSB Chorales dataset.

Contribution

It introduces TonicNet, a novel GRU-based model that predicts chords before individual notes, leveraging feature-rich encoding for enhanced polyphonic music modeling.

Findings

Feature-rich encoding improves model performance

Predicting chords before notes yields better results

Achieved state-of-the-art on JSB Chorales dataset

Abstract

This paper explores sequential modelling of polyphonic music with deep neural networks. While recent breakthroughs have focussed on network architecture, we demonstrate that the representation of the sequence can make an equally significant contribution to the performance of the model as measured by validation set loss. By extracting salient features inherent to the training dataset, the model can either be conditioned on these features or trained to predict said features as extra components of the sequences being modelled. We show that training a neural network to predict a seemingly more complex sequence, with extra features included in the series being modelled, can improve overall model performance significantly. We first introduce TonicNet, a GRU-based model trained to initially predict the chord at a given time-step before then predicting the notes of each voice at that time-step, in contrast with the typical approach of predicting only the notes. We then evaluate TonicNet on the canonical JSB Chorales dataset and obtain state-of-the-art results.