An Architecture for Deep, Hierarchical Generative Models

TL;DR

This paper introduces a deep hierarchical generative model architecture with enhanced connectivity and autoregressive components, enabling effective training of models with over ten layers of latent variables for improved image modeling and interpretability.

Contribution

The paper proposes a novel architecture for deep, hierarchical generative models that incorporates deterministic paths and autoregressive components, allowing end-to-end training of very deep models.

Findings

Achieves state-of-the-art results on image benchmarks

Exposes latent class structures without labels

Provides convincing image imputations

Abstract

We present an architecture which lets us train deep, directed generative models with many layers of latent variables. We include deterministic paths between all latent variables and the generated output, and provide a richer set of connections between computations for inference and generation, which enables more effective communication of information throughout the model during training. To improve performance on natural images, we incorporate a lightweight autoregressive model in the reconstruction distribution. These techniques permit end-to-end training of models with 10+ layers of latent variables. Experiments show that our approach achieves state-of-the-art performance on standard image modelling benchmarks, can expose latent class structure in the absence of label information, and can provide convincing imputations of occluded regions in natural images.