Learning Deep-Latent Hierarchies by Stacking Wasserstein Autoencoders

Abstract

Probabilistic models with hierarchical-latent-variable structures provide state-of-the-art results amongst non-autoregressive, unsupervised density-based models. However, the most common approach to training such models based on Variational Autoencoders (VAEs) often fails to leverage deep-latent hierarchies; successful approaches require complex inference and optimisation schemes. Optimal Transport is an alternative, non-likelihood-based framework for training generative models with appealing theoretical properties, in principle allowing easier training convergence between distributions. In this work we propose a novel approach to training models with deep-latent hierarchies based on Optimal Transport, without the need for highly bespoke models and inference networks. We show that our method enables the generative model to fully leverage its deep-latent hierarchy, avoiding the well known "latent variable collapse" issue of VAEs; therefore, providing qualitatively better sample generations as well as more interpretable latent representation than the original Wasserstein Autoencoder with Maximum Mean Discrepancy divergence.

References in corpus (13)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Auto-Encoding Variational Bayes
• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• Conditional Image Generation with PixelCNN Decoders
• Pixel Recurrent Neural Networks
• Variational Inference with Normalizing Flows
• Auxiliary Deep Generative Models
• Towards Deeper Understanding of Variational Autoencoding Models
• BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling
• From optimal transport to generative modeling: the VEGAN cookbook
• An Architecture for Deep, Hierarchical Generative Models
• On the Latent Space of Wasserstein Auto-Encoders
• Gaussian mixture models with Wasserstein distance