This is code for a handful of artificial toy data sets we created to evaluate (conditional) invertible neural networks and related methods. Each data set provides a prior distribution over data X, and a forward process that maps X to observations Y in a many-to-one manner. The inverse problem of finding X given an observation Y is thus an ambiguous and non-trivial task.
jxzhangjhu / inn_toy_data Goto Github PK
View Code? Open in Web Editor NEWThis project forked from vislearn/inn_toy_data
Code for artificial toy data sets used to evaluate (conditional) invertible neural networks and related methods
License: MIT License