Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting

TL;DR

This paper introduces a likelihood-free importance weighting method to correct bias in learned generative models, improving their sample quality and utility in applications like data augmentation and policy evaluation.

Contribution

It proposes a novel bias correction technique using classifier-based likelihood ratio estimation, enhancing the accuracy of generative models without requiring explicit likelihoods.

Findings

Improves goodness-of-fit metrics for deep generative models

Reduces bias in sample generation

Enhances application performance in data augmentation and policy evaluation

Abstract

A learned generative model often produces biased statistics relative to the underlying data distribution. A standard technique to correct this bias is importance sampling, where samples from the model are weighted by the likelihood ratio under model and true distributions. When the likelihood ratio is unknown, it can be estimated by training a probabilistic classifier to distinguish samples from the two distributions. We employ this likelihood-free importance weighting method to correct for the bias in generative models. We find that this technique consistently improves standard goodness-of-fit metrics for evaluating the sample quality of state-of-the-art deep generative models, suggesting reduced bias. Finally, we demonstrate its utility on representative applications in a) data augmentation for classification using generative adversarial networks, and b) model-based policy evaluation using off-policy data.