Effects of Dataset properties on the training of GANs

TL;DR

This paper investigates how different properties of training datasets influence the training dynamics and stability of GANs, aiming to understand factors that affect their ability to generate high-quality images.

Contribution

It provides experimental insights into how dataset characteristics impact GAN training stability and outcomes, addressing a gap in understanding GAN training behavior.

Findings

Dataset properties significantly affect GAN training stability.

Certain dataset features correlate with mode collapse.

Training outcomes vary with dataset complexity.

Abstract

Generative Adversarial Networks are a new family of generative models, frequently used for generating photorealistic images. The theory promises for the GAN to eventually reach an equilibrium where generator produces pictures indistinguishable for the training set. In practice, however, a range of problems frequently prevents the system from reaching this equilibrium, with training not progressing ahead due to instabilities or mode collapse. This paper describes a series of experiments trying to identify patterns in regard to the effect of the training set on the dynamics and eventual outcome of the training.