This repository enables statistical inference for multivariate geometric extremes as proposed in:
De Monte et al. (2025) Generative modelling of multivariate geometric extremes using normalising flows, arXiv:2505.02957.
The methodology relies on the normflows package and the methodology of Rezende (2020).
This normflows package itself relies on Pytorch from the Python programming language.
If you are familiar with Python programming and setting up work environments, ignore this section and simply take a look at the main.py file and how it calls our geometric-extremes functions.
Otherwise, this section details how to get code running in Visual Studio Code (VSCode), a standard integrated development environment (IDE) in which you can program and execute Python files.
First, install the latest version of Python and of VSCode.
Create a folder for the project in your repositories, named say geometricExtremesNF/. Open VSCode in the project folder by clicking File + Open folder. Create a virtual environment by typing Ctrl + Shift + P and then Python: Create environment....
Select the Python interpreter from the virtual environment by typing Python: Select interpreter....
More information can be found here.
Install the required Python libraries on which our code relies by executing
pip install normflows, pickle, matplotlib, scipy, tqdm
in the VSCode terminal.
You can clone this repository into geometricExtremesNF/. You should now be able to execute the code from main.py.