This repository is the official PyTorch implementation of Reliable Estimation of KL Divergence using a Discriminator in Reproducing Kernel Hilbert Space.
To install requirements: The code has been tested on Python 3.7, PyTorch 1.6.0. The requirements are minimal : numpy, pandas, matplotlib. You could also run the following:
pip install -r requirements.txt
To estimate mutual information, run the following commands. Early commands generate results as pkl file while the last command plots the figure based on previously saved pkl data:
python kl_main.py
python kl_lip_features.py --lambd 0
python kl_lip_features.py --lambd 1e-05
python kl_lip_features.py --lambd 0.001
python kl_lip_features.py --lambd 0.1
python kl_lip_features.py --lambd 1
python plot_method_array_short.py