Why are Modern GANs Poor Density Models?
Lists
Only you see this
Abstract
Modern generative adversarial networks (GANs) generate extremely realistic images and are generally believed to capture the true data distribution. In this work, we evaluate modern GANs as density models and ask whether they can be used for tasks such as outlier detection and generative classification. We find that the performance of state-of-the-art GANs is very poor on these tasks and is often close to (or worse than) random. For instance, a modern GAN that generates remarkably realistic samples when trained on CIFAR10, consistently assigns higher likelihood to flat images than to images from the training set. To try and understand the source of this poor performance, we show that the likelihood that a GAN assigns to an input image is dominated by the quality of the GAN reconstruction. Surprisingly, GANs often fail to reconstruct images from the training set while they are highly effective at reconstructing images outside the distribution. Taken together, our results indicate that modern GANs do not truly learn the underlying distribution, despite the impressive quality of the generated samples.
BibTeX
@misc{friedman2024why,
author = {Roy Friedman and Yair Weiss},
title = {{Why are Modern GANs Poor Density Models?}},
year = {2024},
url = {https://openreview.net/forum?id=nJsfYo3HDy},
howpublished = {OpenReview},
note = {Submitted to ICLR 2024},
}Loading related content...