Discriminator Rejection Sampling

Abstract

We propose a rejection sampling scheme using the discriminator of a GAN to approximately correct errors in the GAN generator distribution. We show that under quite strict assumptions, this will allow us to recover the data distribution exactly. We then examine where those strict assumptions break down and design a practical algorithm - called Discriminator Rejection Sampling (DRS) - that can be used on real data-sets. Finally, we demonstrate the efficacy of DRS on a mixture of Gaussians and on the SAGAN model, state-of-the-art in the image generation task at the time of developing this work. On ImageNet, we train an improved baseline that increases the Inception Score from 52.52 to 62.36 and reduces the Frechet Inception Distance from 18.65 to 14.79. We then use DRS to further improve on this baseline, improving the Inception Score to 76.08 and the FID to 13.75.

Published as a conference paper at ICLR 2019

Cited by in corpus (18)

• Freeze the Discriminator: a Simple Baseline for Fine-Tuning GANs
• Generating Diverse High-Fidelity Images with VQ-VAE-2
• Subsampling Generative Adversarial Networks: Density Ratio Estimation in Feature Space with Softplus Loss
• A RAD approach to deep mixture models
• Residual Energy-Based Models for Text
• Efficient Generation of Structured Objects with Constrained Adversarial Networks
• Self-Diagnosing GAN: Diagnosing Underrepresented Samples in Generative Adversarial Networks
• Private Post-GAN Boosting
• Learning disconnected manifolds: a no GANs land
• A Contrastive Learning Approach for Training Variational Autoencoder Priors
• Generalized Energy Based Models
• Automating Generative Deep Learning for Artistic Purposes: Challenges and Opportunities
• Model Extraction and Defenses on Generative Adversarial Networks
• Effective Data Augmentation with Multi-Domain Learning GANs
• Refining Deep Generative Models via Discriminator Gradient Flow
• A cost-effective method for improving and re-purposing large, pre-trained GANs by fine-tuning their class-embeddings
• Collaborative Sampling in Generative Adversarial Networks
• Reparameterized Sampling for Generative Adversarial Networks