Dropout Induced Noise for Co-Creative GAN Systems

TL;DR

This paper introduces a novel use of Dropout in GANs to produce diverse outputs from a single input, serving as an alternative to latent space exploration especially for constrained input tasks.

Contribution

It presents a new method leveraging Dropout in GANs to generate multiple outputs, addressing limitations of traditional latent space exploration.

Findings

Dropout enables multiple diverse outputs for a single input in GANs.

The method preserves input constraints better than latent space sampling.

Applicable to tasks like A-to-B translation.

Abstract

This paper demonstrates how Dropout can be used in Generative Adversarial Networks to generate multiple different outputs to one input. This method is thought as an alternative to latent space exploration, especially if constraints in the input should be preserved, like in A-to-B translation tasks.