A Hierarchical Latent Vector Model for Learning Long-Term Structure in Music

TL;DR

This paper introduces a hierarchical latent vector model for music sequences that improves long-term structure modeling, sampling, and interpolation by using a hierarchical decoder to better utilize latent representations.

Contribution

It proposes a hierarchical decoder architecture for VAEs that effectively models long-term dependencies in music sequences, addressing posterior collapse.

Findings

Significantly better sampling quality

Improved interpolation between musical sequences

Enhanced reconstruction accuracy

Abstract

The Variational Autoencoder (VAE) has proven to be an effective model for producing semantically meaningful latent representations for natural data. However, it has thus far seen limited application to sequential data, and, as we demonstrate, existing recurrent VAE models have difficulty modeling sequences with long-term structure. To address this issue, we propose the use of a hierarchical decoder, which first outputs embeddings for subsequences of the input and then uses these embeddings to generate each subsequence independently. This structure encourages the model to utilize its latent code, thereby avoiding the "posterior collapse" problem, which remains an issue for recurrent VAEs. We apply this architecture to modeling sequences of musical notes and find that it exhibits dramatically better sampling, interpolation, and reconstruction performance than a "flat" baseline model. An implementation of our "MusicVAE" is available online at http://g.co/magenta/musicvae-code.