Scalable Factorized Hierarchical Variational Autoencoder Training

TL;DR

This paper introduces a scalable hierarchical sampling training algorithm for factorized hierarchical variational autoencoders, enabling effective training on large-scale datasets for improved disentangled representation learning.

Contribution

The paper proposes a novel hierarchical sampling training algorithm that enhances scalability, reduces runtime and memory issues, and improves hyperparameter optimization for FHVAE models.

Findings

Effective training on datasets from 3 to 1,000 hours.

Models demonstrate improved disentanglement and interpretability.

Visualization method aids qualitative evaluation.

Abstract

Deep generative models have achieved great success in unsupervised learning with the ability to capture complex nonlinear relationships between latent generating factors and observations. Among them, a factorized hierarchical variational autoencoder (FHVAE) is a variational inference-based model that formulates a hierarchical generative process for sequential data. Specifically, an FHVAE model can learn disentangled and interpretable representations, which have been proven useful for numerous speech applications, such as speaker verification, robust speech recognition, and voice conversion. However, as we will elaborate in this paper, the training algorithm proposed in the original paper is not scalable to datasets of thousands of hours, which makes this model less applicable on a larger scale. After identifying limitations in terms of runtime, memory, and hyperparameter optimization, we propose a hierarchical sampling training algorithm to address all three issues. Our proposed method is evaluated comprehensively on a wide variety of datasets, ranging from 3 to 1,000 hours and involving different types of generating factors, such as recording conditions and noise types. In addition, we also present a new visualization method for qualitatively evaluating the performance with respect to the interpretability and disentanglement. Models trained with our proposed algorithm demonstrate the desired characteristics on all the datasets.