Data Dieting in GAN Training [article]

Jamal Toutouh, Una-May O'Reilly, Erik Hemberg
2020 arXiv   pre-print
We investigate training Generative Adversarial Networks, GANs, with less data. Subsets of the training dataset can express empirical sample diversity while reducing training resource requirements, e.g. time and memory. We ask how much data reduction impacts generator performance and gauge the additive value of generator ensembles. In addition to considering stand-alone GAN training and ensembles of generator models, we also consider reduced data training on an evolutionary GAN training
more » ... named Redux-Lipizzaner. Redux-Lipizzaner makes GAN training more robust and accurate by exploiting overlapping neighborhood-based training on a spatial 2D grid. We conduct empirical experiments on Redux-Lipizzaner using the MNIST and CelebA data sets.
arXiv:2004.04642v1 fatcat:46sccshdrfgarebf77twxq2yvm