GAN and VAE from an Optimal Transport Point of View
By Aude Genevay, G. Peyr'e, Marco Cuturi. 2017
63 citations recorded by the source corpus.
This short article revisits some of the ideas introduced in arXiv:1701.07875 and arXiv:1705.07642 in a simple setup. This sheds some lights on the connexions between Variational Autoencoders (VAE), Generative Adversarial Networks (GAN) and Minimum Kantorovitch Estimators (MKE).
Subfields: Aerospace Engineering, Generative Models, Mathematical Optimization
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
In Generative Models
- Han Zhang
- Dimitris N. Metaxas
- Tim Salimans
- Paris Perdikaris
- Alec Radford
- Yibo Yang
- Augustus Odena
- Stefano Ermon
In Mathematical Optimization
Citations in the corpus
- Autoregressive Quantile Networks for Generative Modeling (2018, neutral): “The optimal transport metrics W c , for underlying metric c ( x, x (cid:48) ) , and in particular the p -Wasserstein distance, when c is an L p metric, have frequently been proposed as being well-suited replacements to KL (Bousquet et al., 2017; Genevay et al., 2017).”
- Improving GANs Using Optimal Transport (2018, methodology): “Genevay et al. (2017b) then approximate this distance by evaluating it on mini-batches of data X , Y consisting of K data vectors x , y .”
- Conditional deep surrogate models for stochastic, high-dimensional, and multi-fidelity systems (2019, background): “…to further elucidate the capabilities and limitations of such models [31–36], enhance the interpretability of their results [37–39], as well as establish formal connections with well studied topics in mathematics and statistics, including importance sampling [40,41] and optimal transport [42–44].”