Neural Variational Inference and Learning in Belief Networks

TL;DR

This paper introduces a fast, scalable variational inference method using a feedforward network for belief networks, enabling efficient training of complex models on large datasets and outperforming previous algorithms.

Contribution

It presents a novel non-iterative inference approach with variance reduction techniques, improving training efficiency and accuracy for sigmoid belief networks.

Findings

Outperforms wake-sleep algorithm on MNIST

Achieves state-of-the-art results on Reuters RCV1 dataset

Enables scalable training of deep belief networks

Abstract

Highly expressive directed latent variable models, such as sigmoid belief networks, are difficult to train on large datasets because exact inference in them is intractable and none of the approximate inference methods that have been applied to them scale well. We propose a fast non-iterative approximate inference method that uses a feedforward network to implement efficient exact sampling from the variational posterior. The model and this inference network are trained jointly by maximizing a variational lower bound on the log-likelihood. Although the naive estimator of the inference model gradient is too high-variance to be useful, we make it practical by applying several straightforward model-independent variance reduction techniques. Applying our approach to training sigmoid belief networks and deep autoregressive networks, we show that it outperforms the wake-sleep algorithm on MNIST and achieves state-of-the-art results on the Reuters RCV1 document dataset.