Gradient Estimators for Implicit Models

TL;DR

This paper introduces the Stein gradient estimator, a novel method for directly estimating the score function of implicit models, improving training accuracy and sample diversity in applications like GANs and meta-learning.

Contribution

The paper proposes the Stein gradient estimator, enabling direct score function estimation for implicit models, reducing reliance on approximations and enhancing model performance.

Findings

Improved sample diversity in entropy-regularized GANs

Enhanced meta-learning for approximate inference

Demonstrated efficacy of the estimator in various applications

Abstract

Implicit models, which allow for the generation of samples but not for point-wise evaluation of probabilities, are omnipresent in real-world problems tackled by machine learning and a hot topic of current research. Some examples include data simulators that are widely used in engineering and scientific research, generative adversarial networks (GANs) for image synthesis, and hot-off-the-press approximate inference techniques relying on implicit distributions. The majority of existing approaches to learning implicit models rely on approximating the intractable distribution or optimisation objective for gradient-based optimisation, which is liable to produce inaccurate updates and thus poor models. This paper alleviates the need for such approximations by proposing the Stein gradient estimator, which directly estimates the score function of the implicitly defined distribution. The efficacy of the proposed estimator is empirically demonstrated by examples that include meta-learning for approximate inference, and entropy regularised GANs that provide improved sample diversity.