Discriminator optimal transport

Abstract

Within a broad class of generative adversarial networks, we show that discriminator optimization process increases a lower bound of the dual cost function for the Wasserstein distance between the target distribution $p$ and the generator distribution $p_G$. It implies that the trained discriminator can approximate optimal transport (OT) from $p_G$ to $p$.Based on some experiments and a bit of OT theory, we propose a discriminator optimal transport (DOT) scheme to improve generated images. We show that it improves inception score and FID calculated by un-conditional GAN trained by CIFAR-10, STL-10 and a public pre-trained model of conditional GAN by ImageNet.

math errors corrected, note added

Cited by in corpus (9)

• Freeze the Discriminator: a Simple Baseline for Fine-Tuning GANs
• LOGAN: Latent Optimisation for Generative Adversarial Networks
• Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling
• Score-based Generative Modeling in Latent Space
• Alternating the Population and Control Neural Networks to Solve High-Dimensional Stochastic Mean-Field Games
• Generalized Energy Based Models
• Refining Deep Generative Models via Discriminator Gradient Flow
• MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining
• Reparameterized Sampling for Generative Adversarial Networks