Adaptive Density Estimation for Generative Models - Archive ouverte HAL
Conference Papers Year : 2019

Adaptive Density Estimation for Generative Models

Abstract

Unsupervised learning of generative models has seen tremendous progress over recent years, in particular due to generative adversarial networks (GANs), variational autoencoders, and flow-based models. GANs have dramatically improved sample quality, but suffer from two drawbacks: (i) they mode-drop, i.e., do not cover the full support of the train data, and (ii) they do not allow for likelihood evaluations on held-out data. In contrast likelihood-based training encourages models to cover the full support of the train data, but yields poorer samples. These mutual short comings can in principle be addressed by training generative latent variable models in a hybrid adversarial-likelihood manner. However, we show that commonly made parametric assumptions create a conflict between them, making successful hybrid models non trivial. As a solution, we propose to use deep invertible transformations in the latent variable decoder. This approach allows for likelihood computations in image space, is more efficient than fully invertible models, and can take full advantage of adversarial training. We show that our model significantly improves over existing hybrid models: offering GAN-like samples, IS and FID scores thatare competitive with fully adversarial models, and improved likelihood scores.
Fichier principal
Vignette du fichier
main.pdf (8.15 Mo) Télécharger le fichier
Origin Files produced by the author(s)
Loading...

Dates and versions

hal-01886285 , version 1 (02-10-2018)
hal-01886285 , version 2 (01-03-2019)
hal-01886285 , version 3 (03-12-2019)
hal-01886285 , version 4 (03-01-2020)
hal-01886285 , version 5 (14-02-2020)

Identifiers

  • HAL Id : hal-01886285 , version 5

Cite

Thomas Lucas, Konstantin Shmelkov, Cordelia Schmid, Karteek Alahari, Jakob Verbeek. Adaptive Density Estimation for Generative Models. NeurIPS 2019 - Thirty-third Conference on Neural Information Processing Systems, Dec 2019, Vancouver, Canada. pp.11993-12003. ⟨hal-01886285v5⟩
1211 View
1297 Download

Share

More