This repository contains reproducible code for the manuscript Constructing Large Nonstationary Spatio-Temporal Covariance Models via Compositional Warpings by Quan Vu, Andrew Zammit-Mangion and Stephen J. Chuter.
To reproduce the results and the figures in the manuscript, please download this repository.
Ensure the required packages are installed. Please ensure to set up a conda environment with TensorFlow v1.15 to run the TensorFlow code, and modify the file scripts/utils.R to load the correct environment.
The code to reproduce the results of the manuscript is saved in the folder scripts/, and the code to reproduce the figures is saved in the folder scripts_plot/. Once the repository is downloaded, run the bash file run.sh to reproduce the results in the manuscript. To reproduce the figures in the manuscript, please run the bash file run_plot.sh. The results in the manuscript are saved in the folders results/ while the figures used in the manuscript are saved in the folder figures/.