Wasserstein GANs Work Because They Fail (to Approximate the Wasserstein Distance)
read the original abstract
Wasserstein GANs are based on the idea of minimising the Wasserstein distance between a real and a generated distribution. We provide an in-depth mathematical analysis of differences between the theoretical setup and the reality of training Wasserstein GANs. In this work, we gather both theoretical and empirical evidence that the WGAN loss is not a meaningful approximation of the Wasserstein distance. Moreover, we argue that the Wasserstein distance is not even a desirable loss function for deep generative models, and conclude that the success of Wasserstein GANs can in truth be attributed to a failure to approximate the Wasserstein distance.
This paper has not been read by Pith yet.
Forward citations
Cited by 4 Pith papers
-
Ensemble Distributionally Robust Bayesian Optimisation with Continuous Context
A tractable ensemble distributionally robust Bayesian optimization method achieves improved sublinear regret bounds under context uncertainty.
-
Ensemble Distributionally Robust Bayesian Optimisation with Continuous Context
EDRBO uses ensemble surrogates and Wasserstein ambiguity sets to robustify BO acquisition functions against context distribution mismatch, with sublinear regret O(γ_T √T) and SOTA empirical results on continuous contexts.
-
Learning to Emulate Chaos: Adversarial Optimal Transport Regularization
Adversarial optimal transport objectives jointly learn summary statistics and a chaotic-system emulator from a single noisy trajectory, improving long-term statistical fidelity over handcrafted-feature baselines.
-
Learning to Emulate Chaos: Adversarial Optimal Transport Regularization
Adversarial optimal transport objectives train neural emulators with improved long-term statistical fidelity on chaotic systems.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.