CS-He / torchKRnet

This is an unofficial implementation of the KRnet with Pytorch, which Tensorflow originally implemented.

KRnet implemented by Pytorch

unofficial implementation Code for KRnet https://doi.org/10.1016/j.jcp.2022.111080.

Install

Pytorch needs to be installed independently following the instruction on the Pytorch homepage. This repo also has dependency on numpy, matplotlib, sklearn.

Experiments

Experiments may be run with the following commands. Run the following script for eight Gaussian density estimation.

python density_estimation.py # density estimation for eight Gaussian

If you successfully run this script, you will get the result image like that.

Citation

If you use this code in your research, please cite it as:

@article{tang2020deep, title={Deep density estimation via invertible block-triangular mapping}, volume={10}, ISSN={2095-0349}, url={http://dx.doi.org/10.1016/j.taml.2020.01.023}, DOI={10.1016/j.taml.2020.01.023}, number={3}, journal={Theoretical and Applied Mechanics Letters}, publisher={Elsevier BV}, author={Tang, Keju and Wan, Xiaoliang and Liao, Qifeng}, year={2020}, month={Mar}, pages={143–148} }

and

@article{tang2022adaptive, title={Adaptive deep density approximation for Fokker-Planck equations}, volume={457}, ISSN={0021-9991}, url={http://dx.doi.org/10.1016/j.jcp.2022.111080}, DOI={10.1016/j.jcp.2022.111080}, journal={Journal of Computational Physics}, publisher={Elsevier BV}, author={Tang, Kejun and Wan, Xiaoliang and Liao, Qifeng}, year={2022}, month={May}, pages={111080} }