Multimodal Transitions for Generative Stochastic Networks

TL;DR

This paper introduces multimodal transition distributions for Generative Stochastic Networks using NADE models, enabling better modeling of complex data distributions and improving sample quality.

Contribution

It is the first to incorporate multimodal transition distributions in GSNs, leveraging NADE models for enhanced generative capabilities.

Findings

Multimodal GSNs outperform unimodal GSNs in experiments.

NADE-based transition models effectively capture complex data distributions.

Multimodal transitions improve sample diversity and quality.

Abstract

Generative Stochastic Networks (GSNs) have been recently introduced as an alternative to traditional probabilistic modeling: instead of parametrizing the data distribution directly, one parametrizes a transition operator for a Markov chain whose stationary distribution is an estimator of the data generating distribution. The result of training is therefore a machine that generates samples through this Markov chain. However, the previously introduced GSN consistency theorems suggest that in order to capture a wide class of distributions, the transition operator in general should be multimodal, something that has not been done before this paper. We introduce for the first time multimodal transition distributions for GSNs, in particular using models in the NADE family (Neural Autoregressive Density Estimator) as output distributions of the transition operator. A NADE model is related to an RBM (and can thus model multimodal distributions) but its likelihood (and likelihood gradient) can be computed easily. The parameters of the NADE are obtained as a learned function of the previous state of the learned Markov chain. Experiments clearly illustrate the advantage of such multimodal transition distributions over unimodal GSNs.