Dropout Induced Noise for Co-Creative GAN Systems
This addresses a need for co-creative systems in AI art or design, but it appears incremental as it adapts an existing technique (Dropout) to a specific GAN application.
The paper tackles the problem of generating multiple diverse outputs from a single input in GANs, particularly for tasks like A-to-B translation where input constraints must be preserved, by using Dropout to induce noise, resulting in a method that produces varied outputs without altering the latent space.
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.