f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
This work provides a more general framework for training generative models, benefiting researchers in machine learning by extending beyond adversarial methods, though it is incremental as it builds on existing variational divergence estimation.
The paper tackles the problem of training generative neural samplers, which cannot compute likelihoods, by showing that generative-adversarial training is a special case of variational divergence estimation and that any f-divergence can be used for this purpose, leading to improved training and model quality.
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.