Imposing higher-level Structure in Polyphonic Music Generation using Convolutional Restricted Boltzmann Machines and Constraints

TL;DR

This paper presents a novel method combining Convolutional Restricted Boltzmann Machines with constraint optimization and simulated annealing to generate polyphonic music with controllable higher-level structures like self-similarity, meter, and tonality.

Contribution

It introduces a new approach that integrates generative modeling with constraint-based control to impose higher-level musical structures during generation.

Findings

Controlled higher-level musical features achieved

Preserved local musical coherence

Effective transfer of structural properties from template

Abstract

We introduce a method for imposing higher-level structure on generated, polyphonic music. A Convolutional Restricted Boltzmann Machine (C-RBM) as a generative model is combined with gradient descent constraint optimisation to provide further control over the generation process. Among other things, this allows for the use of a "template" piece, from which some structural properties can be extracted, and transferred as constraints to the newly generated material. The sampling process is guided with Simulated Annealing to avoid local optima, and to find solutions that both satisfy the constraints, and are relatively stable with respect to the C-RBM. Results show that with this approach it is possible to control the higher-level self-similarity structure, the meter, and the tonal properties of the resulting musical piece, while preserving its local musical coherence.