Fostering Diversity in Spatial Evolutionary Generative Adversarial Networks

Generative adversary networks (GANs) suffer from training pathologies such as\ninstability and mode collapse, which mainly arise from a lack of diversity in\ntheir adversarial interactions. Co-evolutionary GAN (CoE-GAN) training\nalgorithms have shown to be resilient to these pathologies. This article\nintroduces Mustangs, a spatially distributed CoE-GAN, which fosters diversity\nby using different loss functions during the training. Experimental analysis on\nMNIST and CelebA demonstrated that Mustangs trains statistically more accurate\ngenerators.\n