Learning to Draw Samples: With Application to Amortized MLE for Generative Adversarial Learning

TL;DR

This paper introduces a simple, flexible algorithm for training neural networks to generate samples from complex distributions, applicable to various models including deep energy models, with competitive results in image generation.

Contribution

The paper presents a novel Stein variational gradient-based method for training neural samplers that can be applied to any differentiable black-box model and introduces an amortized MLE approach for deep energy models.

Findings

Effective sampling from unnormalized distributions

Competitive image generation quality

Versatile application to deep energy models

Abstract

We propose a simple algorithm to train stochastic neural networks to draw samples from given target distributions for probabilistic inference. Our method is based on iteratively adjusting the neural network parameters so that the output changes along a Stein variational gradient that maximumly decreases the KL divergence with the target distribution. Our method works for any target distribution specified by their unnormalized density function, and can train any black-box architectures that are differentiable in terms of the parameters we want to adapt. As an application of our method, we propose an amortized MLE algorithm for training deep energy model, where a neural sampler is adaptively trained to approximate the likelihood function. Our method mimics an adversarial game between the deep energy model and the neural sampler, and obtains realistic-looking images competitive with the state-of-the-art results.