Improved Procedures for Training Primal Wasserstein GANs

Primal Wasserstein GANs are a variant of Generative Adversarial Networks (i.e., GANs), which optimize the primal form of empirical Wasserstein distance directly. However, the high computational complexity and training instability are the main challenges of this framework. Accordingly, to address these problems, we propose several procedures for improving the training of Primal Wasserstein GANs. We test the effectiveness of our proposed procedures on MNIST, CIFAR-10, LSUN-Bedroom and ImageNet-Dog category datasets, and the extensive experimental results confirm that our method is capable of generating high-quality images and obtaining high inception score. Importantly, we demonstrate that our method is more time efficient compared with other generative model techniques.