Potential Flow Generator with $L_2$ Optimal Transport Regularity for Generative Models

TL;DR

This paper introduces a potential flow generator with $L_2$ optimal transport regularity that enhances various generative models, demonstrating robustness and effectiveness in image translation tasks with unpaired data.

Contribution

The paper presents a novel potential flow generator with $L_2$ optimal transport regularity, compatible with multiple generative models, improving robustness and translation quality.

Findings

Effective in 2D problems demonstrating correctness and robustness.

Improves image translation quality with unpaired MNIST and CelebA datasets.

Abstract

We propose a potential flow generator with $L_2$ optimal transport regularity, which can be easily integrated into a wide range of generative models including different versions of GANs and flow-based models. We show the correctness and robustness of the potential flow generator in several 2D problems, and illustrate the concept of "proximity" due to the $L_2$ optimal transport regularity. Subsequently, we demonstrate the effectiveness of the potential flow generator in image translation tasks with unpaired training data from the MNIST dataset and the CelebA dataset.