Investigating Under and Overfitting in Wasserstein Generative Adversarial Networks
This addresses the problem of understanding and mitigating training instabilities in GANs for researchers and practitioners, but it is incremental as it builds on existing Wasserstein GAN frameworks.
The study investigated under and overfitting in Wasserstein GANs by using unseen discriminators to measure generalization, finding that discriminator capacity significantly affects generator quality and that poor generator performance aligns with discriminator underfitting, with no evidence of overfitting observed on datasets like CIFAR10, CIFAR100, and CelebA.
We investigate under and overfitting in Generative Adversarial Networks (GANs), using discriminators unseen by the generator to measure generalization. We find that the model capacity of the discriminator has a significant effect on the generator's model quality, and that the generator's poor performance coincides with the discriminator underfitting. Contrary to our expectations, we find that generators with large model capacities relative to the discriminator do not show evidence of overfitting on CIFAR10, CIFAR100, and CelebA.