Cycle-consistent adversarial learning as approximate bayesian inference

Publication
arXiv preprint arXiv:1806.01771