Discriminator optimal transport

About

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.

Akinori Tanaka · 2019

Related benchmarks

Task	Dataset	Result
Image Generation	CIFAR-10	Inception Score8.02	178
Image Generation	ImageNet	FID—	106
Image Generation	STL-10 (test)	Inception Score—	59
Unconditional Image Generation	CIFAR10	BPD—	33
Generative Modeling	CIFAR10 (test)	FID15.78	19
Image Generation	STL10	Inception Score9.45	15
Image Generation	CIFAR-10 pixel space (test)	FID15.78	7
Generative Modeling	25Gaussians	High Quality (%)0.698	6