Demonstrating the Evolution of GANs through t-SNE

TL;DR

This paper introduces a t-SNE based evaluation method to visualize and assess the training progress of GANs, demonstrating how evolutionary algorithms like COEGAN improve model quality and mitigate issues like mode collapse.

Contribution

It proposes a novel t-SNE based metric using discriminator features to evaluate GAN training progress and visualizes the evolution of GANs trained with COEGAN.

Findings

t-SNE visualizations show progressive improvement in generator outputs

The proposed metric correlates with visual quality and mode coverage

COEGAN effectively reduces mode collapse over generations

Abstract

Generative Adversarial Networks (GANs) are powerful generative models that achieved strong results, mainly in the image domain. However, the training of GANs is not trivial, presenting some challenges tackled by different strategies. Evolutionary algorithms, such as COEGAN, were recently proposed as a solution to improve the GAN training, overcoming common problems that affect the model, such as vanishing gradient and mode collapse. In this work, we propose an evaluation method based on t-distributed Stochastic Neighbour Embedding (t-SNE) to assess the progress of GANs and visualize the distribution learned by generators in training. We propose the use of the feature space extracted from trained discriminators to evaluate samples produced by generators and from the input dataset. A metric based on the resulting t-SNE maps and the Jaccard index is proposed to represent the model quality. Experiments were conducted to assess the progress of GANs when trained using COEGAN. The results show both by visual inspection and metrics that the Evolutionary Algorithm gradually improves discriminators and generators through generations, avoiding problems such as mode collapse.