# Investigating Under and Overfitting in Wasserstein Generative Adversarial Networks

arXiv: https://arxiv.org/abs/1910.14137 · HTML: https://mlsanity.com/paper/1910.14137/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 30 Oct 2019.

## Built on by (most cited first, 7 of 7)

- [How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Models](https://mlsanity.com/paper/2102.08921/) (2021)
- [Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samples](https://mlsanity.com/paper/2302.04440/) (2023)
- [Reducing Training Sample Memorization in GANs by Training with Memorization Rejection](https://mlsanity.com/paper/2210.12231/) (2022)
- [Generative Minimization Networks: Training GANs Without Competition](https://mlsanity.com/paper/2103.12685/) (2021)
- [Steering Language Generation: Harnessing Contrastive Expert Guidance and Negative Prompting for Coherent and Diverse Synthetic Data Generation](https://mlsanity.com/paper/2308.07645/) (2023)
- [Bidirectional Adversarial Autoencoders for the design of Plasmonic Metasurfaces](https://mlsanity.com/paper/2405.04056/) (2024)
- [Provably Near-Optimal Federated Ensemble Distillation with Negligible Overhead](https://mlsanity.com/paper/2502.06349/) (2025)

## Abstract (arXiv)

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.