GAN and VAE from an Optimal Transport Point of View

06/06/2017

∙

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).

READ FULL TEXT

GAN and VAE from an Optimal Transport Point of View

Implicit Discriminator in Variational Autoencoder

Conditional Autoencoders with Adversarial Information Factorization

Adversarial Networks and Autoencoders: The Primal-Dual Relationship and Generalization Bounds

A Probe Towards Understanding GAN and VAE Models

Top-N: Equivariant set and graph generation without exchangeability

Information bottleneck through variational glasses

InfoGAN-CR: Disentangling Generative Adversarial Networks with Contrastive Regularizers

GAN and VAE from an Optimal Transport Point of View

Related Research

Implicit Discriminator in Variational Autoencoder

Conditional Autoencoders with Adversarial Information Factorization

Adversarial Networks and Autoencoders: The Primal-Dual Relationship and Generalization Bounds

A Probe Towards Understanding GAN and VAE Models

Top-N: Equivariant set and graph generation without exchangeability

Information bottleneck through variational glasses

InfoGAN-CR: Disentangling Generative Adversarial Networks with Contrastive Regularizers