Comparing Probabilistic Models for Melodic Sequences

TL;DR

This paper compares two probabilistic models, Dirichlet-VMM and TC-RBM, for modeling melodic sequences within a music genre, demonstrating their superior performance over the state-of-the-art VMM in prediction and statistical matching.

Contribution

It introduces a comparative analysis of Dirichlet-VMM and TC-RBM models for melodic sequence modeling, highlighting their ability to learn musical features and outperform existing models.

Findings

Both models outperform the VMM in prediction accuracy.

Dirichlet-VMM marginally outperforms TC-RBM.

Models better match the statistical properties of the music genre.

Abstract

Modelling the real world complexity of music is a challenge for machine learning. We address the task of modeling melodic sequences from the same music genre. We perform a comparative analysis of two probabilistic models; a Dirichlet Variable Length Markov Model (Dirichlet-VMM) and a Time Convolutional Restricted Boltzmann Machine (TC-RBM). We show that the TC-RBM learns descriptive music features, such as underlying chords and typical melody transitions and dynamics. We assess the models for future prediction and compare their performance to a VMM, which is the current state of the art in melody generation. We show that both models perform significantly better than the VMM, with the Dirichlet-VMM marginally outperforming the TC-RBM. Finally, we evaluate the short order statistics of the models, using the Kullback-Leibler divergence between test sequences and model samples, and show that our proposed methods match the statistics of the music genre significantly better than the VMM.