# Improving Polyphonic Music Models with Feature-Rich Encoding

arXiv: https://arxiv.org/abs/1911.11775 · HTML: https://mlsanity.com/paper/1911.11775/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Nov 2019.

## Built on by (most cited first, 6 of 6)

- [Music Composition with Deep Learning: A Review](https://mlsanity.com/paper/2108.12290/) (2021)
- [Choir Transformer: Generating Polyphonic Music with Relative Attention on Transformer](https://mlsanity.com/paper/2308.02531/) (2023)
- [From Words to Music: A Study of Subword Tokenization Techniques in Symbolic Music Generation](https://mlsanity.com/paper/2304.08953/) (2023)
- [ProGress: Structured Music Generation via Graph Diffusion and Hierarchical Music Analysis](https://mlsanity.com/paper/2510.10249/) (2025)
- [CoCoFormer: A controllable feature-rich polyphonic music generation method](https://mlsanity.com/paper/2310.09843/) (2023)
- [A Novel Bi-LSTM And Transformer Architecture For Generating Tabla Music](https://mlsanity.com/paper/2404.05765/) (2024)

## Abstract (arXiv)

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.