A Sparsity-promoting Dictionary Model for Variational Autoencoders

TL;DR

This paper introduces a sparsity-promoting dictionary model for VAEs that enhances interpretability and expressiveness without sacrificing reconstruction quality, using a simple, efficient, and tuning-free approach.

Contribution

It proposes a novel sparsity-promoting dictionary model for VAEs with a learnable Gaussian prior, improving interpretability and sparsity without degrading output quality.

Findings

Outperforms competing methods in speech generation tasks

Promotes sparsity without reducing speech quality

Uses a computationally efficient, tuning-free inference scheme

Abstract

Structuring the latent space in probabilistic deep generative models, e.g., variational autoencoders (VAEs), is important to yield more expressive models and interpretable representations, and to avoid overfitting. One way to achieve this objective is to impose a sparsity constraint on the latent variables, e.g., via a Laplace prior. However, such approaches usually complicate the training phase, and they sacrifice the reconstruction quality to promote sparsity. In this paper, we propose a simple yet effective methodology to structure the latent space via a sparsity-promoting dictionary model, which assumes that each latent code can be written as a sparse linear combination of a dictionary's columns. In particular, we leverage a computationally efficient and tuning-free method, which relies on a zero-mean Gaussian latent prior with learnable variances. We derive a variational inference scheme to train the model. Experiments on speech generative modeling demonstrate the advantage of the proposed approach over competing techniques, since it promotes sparsity while not deteriorating the output speech quality.