Code for the paper "Tackling the Accuracy-Interpretability Trade-off in a Hierarchy of Machine Learning Models for the Prediction of Extreme Heatwaves"

Useful links

repo of manuscript repo of supplementary material

Preprint of paper:

Setup

To be able to run the notebooks in this repository, you need to clone the Climate-Learning repository.

To do put yoursel in the same directory of this file and run

git clone --recursive https://github.com/georgemilosh/Climate-Learning.git

Reproducing the figures will use only the submodule general_purpose, but training the neural networks and visualizing them needs the full Climate-Learning framework.

Contents of this repo

This repo contains data and notebooks to reproduce the figures and tables presented in our paper. The best way to navigate it is through its notebooks: inside each of them you'll find some explanatory markdown text.

List of notebooks

• Data normalization procedure here
• Performance of the hierarachy here
• Interpretability
  • GA and IINN here
  • CNN
    • Expected Gradient Feature Importance maps here
    • Optimal Input Maps here
  • Scatnet here

Debugging help

• Test if you can properly load our neural networks here