This paper proposes a set of methods that enables low precision posit™ arithmetic to be
successfully used for the training of generative adversarial networks (GANs) with minimal …