# Evaluating the distribution learning capabilities of GANs

## TL;DR

This paper assesses how well GANs can learn different types of data distributions, revealing limitations in capturing discontinuities, object counts, and the balance between generalization and learning.

## Contribution

It provides a systematic evaluation of GANs on synthetic datasets, highlighting specific weaknesses in distribution learning capabilities.

## Key findings

- GANs struggle with discontinuous point distributions
- GANs do not effectively learn object counts in images
- There is a tension between generalization and learning in GANs

## Abstract

We evaluate the distribution learning capabilities of generative adversarial networks by testing them on synthetic datasets. The datasets include common distributions of points in $R^n$ space and images containing polygons of various shapes and sizes. We find that by and large GANs fail to faithfully recreate point datasets which contain discontinous support or sharp bends with noise. Additionally, on image datasets, we find that GANs do not seem to learn to count the number of objects of the same kind in an image. We also highlight the apparent tension between generalization and learning in GANs.