Improved BiGAN training with marginal likelihood equalization

Abstract

We propose a novel training procedure for improving the performance of generative adversarial networks (GANs), especially to bidirectional GANs. First, we enforce that the empirical distribution of the inverse inference network matches the prior distribution, which favors the generator network reproducibility on the seen samples. Second, we have found that the marginal log-likelihood of the samples shows a severe overrepresentation of a certain type of samples. To address this issue, we propose to train the bidirectional GAN using a non-uniform sampling for the mini-batch selection, resulting in improved quality and variety in generated samples measured quantitatively and by visual inspection. We illustrate our new procedure with the well-known CIFAR10, Fashion MNIST and CelebA datasets.

References in corpus (12)

• Auto-Encoding Variational Bayes
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Spectral Normalization for Generative Adversarial Networks
• BEGAN: Boundary Equilibrium Generative Adversarial Networks
• Learning to Discover Cross-Domain Relations with Generative Adversarial Networks
• Adversarially Learned Inference
• VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learning
• MMD GAN: Towards Deeper Understanding of Moment Matching Network
• Unrolled Generative Adversarial Networks
• Learning in Implicit Generative Models
• Entropic GANs meet VAEs: A Statistical Approach to Compute Sample Likelihoods in GANs
• Understanding the (un)interpretability of natural image distributions using generative models