Data-Dependent Conditional Priors for Unsupervised Learning of Multimodal Data

One of the major shortcomings of variational autoencoders is the inability to produce generations from the individual modalities of data originating from mixture distributions. This is primarily due to the use of a simple isotropic Gaussian as the prior for the latent code in the ancestral sampling...

Full description

Bibliographic Details
Main Authors: Frantzeska Lavda, Magda Gregorová, Alexandros Kalousis
Format: Article
Language:English
Published: MDPI AG 2020-08-01
Series:Entropy
Subjects:
VAE
Online Access:https://www.mdpi.com/1099-4300/22/8/888