An Architecture for Deep, Hierarchical Generative Models

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.

Published in NIPS 2016

Cited by in corpus (11)

• Neural Language Generation: Formulation, Methods, and Evaluation
• Progressive Learning and Disentanglement of Hierarchical Representations
• MHVAE: a Human-Inspired Deep Hierarchical Generative Model for Multimodal Representation Learning
• Generative Hierarchical Models for Parts, Objects, and Scenes
• A Stable Variational Autoencoder for Text Modelling
• Unsupervised Star Galaxy Classification with Cascade Variational Auto-Encoder
• Semi-Implicit Stochastic Recurrent Neural Networks
• Self-Reflective Variational Autoencoder
• Variational Predictive Routing with Nested Subjective Timescales
• Kanerva++: extending The Kanerva Machine with differentiable, locally block allocated latent memory
• Learning Deep-Latent Hierarchies by Stacking Wasserstein Autoencoders