Learning Priors for Adversarial Autoencoders

Hui-PoWang, Wei-Jan Ko, Wen-Hsiao Peng

Feb 11, 2018 (modified: Jun 04, 2018) ICLR 2018 Workshop Submission readers: everyone Show Bibtex
  • Abstract: Recent studies show that the choice of the prior has a profound effect on the expressiveness of deep latent factor models. In this paper, we propose to learn the prior from data for adversarial autoencoders (AAEs). We introduce the notion of code generators to transform manually selected simple priors into ones that can better characterize the data distribution.
  • Keywords: adversarial autoencoder, generative adversarial networks, prior, disentangled representations
  • TL;DR: Learning a better prior from data for adversarial autoencoders