Improved generator objectives for GANs

Abstract

We present a framework to understand GAN training as alternating density ratio estimation and approximate divergence minimization. This provides an interpretation for the mismatched GAN generator and discriminator objectives often used in practice, and explains the problem of poor sample diversity. We also derive a family of generator objectives that target arbitrary $f$-divergences without minimizing a lower bound, and use them to train generative image models that target either improved sample quality or greater sample diversity.

NIPS 2016 Workshop on Adversarial Training

References in corpus (7)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Adversarially Learned Inference
• f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
• Learning in Implicit Generative Models
• On distinguishability criteria for estimating generative models
• Generative Adversarial Nets from a Density Ratio Estimation Perspective

Cited by in corpus (23)

• Improved Training of Wasserstein GANs
• BEGAN: Boundary Equilibrium Generative Adversarial Networks
• Generative Adversarial Networks in Computer Vision: A Survey and Taxonomy
• Learning in Implicit Generative Models
• Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks
• Do GANs actually learn the distribution? An empirical study
• From optimal transport to generative modeling: the VEGAN cookbook
• Stabilizing Generative Adversarial Networks: A Survey
• Solution of Physics-based Bayesian Inverse Problems with Deep Generative Priors
• Few-Shot Adaptation of Generative Adversarial Networks
• Variational f-divergence Minimization
• Improved Training with Curriculum GANs
• Non-saturating GAN training as divergence minimization
• Generate High Resolution Images With Generative Variational Autoencoder
• Expected Information Maximization: Using the I-Projection for Mixture Density Estimation
• MR-GAN: Manifold Regularized Generative Adversarial Networks
• Properties of f-divergences and f-GAN training
• Bidirectional Generative Modeling Using Adversarial Gradient Estimation
• Deep Convolutional GANs for Car Image Generation
• Semi-Implicit Generative Model
• Text Generation with Deep Variational GAN
• Linear Discriminant Generative Adversarial Networks
• Human Annotations Improve GAN Performances