This repo contains an implementation of https://arxiv.org/abs/2007.04873
Changes have been made to obtain the results of the toy_data experiments present in the article:
- Added LinearLU
- Added ActNorm
Mutiple runs of this implementation with different values of centralizing_loss and learning_rate can be found here: https://wandb.ai/mvalente/toy_data_zf/sweeps/v4eewmtw?workspace=user-mvalente
Contains code from https://github.com/SamGalanakis/FlowCompare