Sha, L.; Lukasiewicz, T. Multi-Type Disentanglement Without Adversarial Training. AAAI 2021, 35, 9515-9523.