Investigating Under and Overfitting in Wasserstein Generative Adversarial Networks

2019/10/30 by Ben Adlam, Adlam, Ben, Charles Weill +3 · 2 citations
Computer Science · Decision Sciences · Mathematics · #Advanced Image Processing Techniques #Adversarial Robustness in Machine Learning #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Probabilistic and Robust Engineering Design #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1910.14137

arxiv created 2019/10/30 · openalex publication_date 2019/10/30 · arxiv updated 2019/11/01 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28

Abstract

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.

Citations

• NIPS 2016 Tutorial: Generative Adversarial Networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis

Cited by

• On Filter Generalization for Music Bandwidth Extension Using Deep Neural Networks
• Generative Minimization Networks: Training GANs Without Competition

Related

• Approximability of Discriminators Implies Diversity in GANs
• Generalization Error of GAN from the Discriminator's Perspective
• Approximating Probability Distributions by using Wasserstein Generative Adversarial Networks
• Rethinking Multidimensional Discriminator Output for Generative Adversarial Networks
• Double Descent and Other Interpolation Phenomena in GANs