Timezone: »

 
Poster
Asymptotic Guarantees for Learning Generative Models with the Sliced-Wasserstein Distance
Kimia Nadjahi · Alain Durmus · Umut Simsekli · Roland Badeau

Thu Dec 12 05:00 PM -- 07:00 PM (PST) @ East Exhibition Hall B + C #226
in Theory -- Large Deviations and Asymptotic Analysis »
Minimum expected distance estimation (MEDE) algorithms have been widely used for probabilistic models with intractable likelihood functions and they have become increasingly popular due to their use in implicit generative modeling (e.g.\ Wasserstein generative adversarial networks, Wasserstein autoencoders). Emerging from computational optimal transport, the Sliced-Wasserstein (SW) distance has become a popular choice in MEDE thanks to its simplicity and computational benefits. While several studies have reported empirical success on generative modeling with SW, the theoretical properties of such estimators have not yet been established. In this study, we investigate the asymptotic properties of estimators that are obtained by minimizing SW. We first show that convergence in SW implies weak convergence of probability measures in general Wasserstein spaces. Then we show that estimators obtained by minimizing SW (and also an approximate version of SW) are asymptotically consistent. We finally prove a central limit theorem, which characterizes the asymptotic distribution of the estimators and establish a convergence rate of $\sqrt{n}$, where $n$ denotes the number of observed data points. We illustrate the validity of our theory on both synthetic data and neural networks.
Keywords: Large Deviations and Asymptotic Analysis · Generative Models · Theory · Deep Learning
[ Paper]  [ Poster]  [ Slides] 

Author Information

Kimia Nadjahi (Télécom ParisTech)
Alain Durmus (ENS Paris Saclay)
Umut Simsekli (Institut Polytechnique de Paris/ University of Oxford)
Roland Badeau (Télécom ParisTech)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors