GSNs : Generative Stochastic Networks

TL;DR

This paper introduces Generative Stochastic Networks (GSNs), a new probabilistic modeling framework that learns transition operators of Markov chains, enabling easier training and sampling, validated through experiments on image datasets.

Contribution

The paper proposes GSNs as an alternative to maximum likelihood, providing a new training principle and theoretical justification for related models like denoising auto-encoders.

Findings

GSNs can effectively model data distributions with fewer modes.

They enable training with backpropagation without layerwise pretraining.

Experimental results validate the theoretical advantages on image datasets.

Abstract

We introduce a novel training principle for probabilistic models that is an alternative to maximum likelihood. The proposed Generative Stochastic Networks (GSN) framework is based on learning the transition operator of a Markov chain whose stationary distribution estimates the data distribution. Because the transition distribution is a conditional distribution generally involving a small move, it has fewer dominant modes, being unimodal in the limit of small moves. Thus, it is easier to learn, more like learning to perform supervised function approximation, with gradients that can be obtained by back-propagation. The theorems provided here generalize recent work on the probabilistic interpretation of denoising auto-encoders and provide an interesting justification for dependency networks and generalized pseudolikelihood (along with defining an appropriate joint distribution and sampling mechanism, even when the conditionals are not consistent). We study how GSNs can be used with missing inputs and can be used to sample subsets of variables given the rest. Successful experiments are conducted, validating these theoretical results, on two image datasets and with a particular architecture that mimics the Deep Boltzmann Machine Gibbs sampler but allows training to proceed with backprop, without the need for layerwise pretraining.