f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization

TL;DR

This paper introduces f-GAN, a flexible framework for training generative neural samplers using various divergence measures, extending the adversarial training approach to a broader class of divergences for improved model quality.

Contribution

It generalizes the GAN training method by incorporating any f-divergence, providing a unified variational divergence minimization framework for generative neural models.

Findings

Any f-divergence can be used for training generative neural samplers.

Different divergence choices affect training complexity and model quality.

The approach unifies and extends existing adversarial training methods.

Abstract

Generative neural samplers are probabilistic models that implement sampling using feedforward neural networks: they take a random input vector and produce a sample from a probability distribution defined by the network weights. These models are expressive and allow efficient computation of samples and derivatives, but cannot be used for computing likelihoods or for marginalization. The generative-adversarial training method allows to train such models through the use of an auxiliary discriminative neural network. We show that the generative-adversarial approach is a special case of an existing more general variational divergence estimation approach. We show that any f-divergence can be used for training generative neural samplers. We discuss the benefits of various choices of divergence functions on training complexity and the quality of the obtained generative models.