Code for the implementation of the paper: Embedded-model flows: Combining the inductive biases of model-free deep learning and explicit probabilistic modeling. This repo contains all the code used in the experiments. The code is currently under refactoring for camera ready submission. In the meantime, please refer to an example usage for the time series toy experiments in this notebook.
Some scripts refer to the Embedded Model Flows as Normalizing Programs and Gated Normalizing Programs, as that was the original name used for such models. The names will be refactored before the official release.