Studying the Generalization Behavior of Surrogate Models for Punch-Bending by Generating Plausible Counterfactuals

This repository contains scripts and the data to reproduced the results for the paper:

@article{mahajan2019preserving,
  title        = {Studying the Generalization Behavior of Surrogate Models for Punch-Bending by Generating Plausible Counterfactuals},
  author       = {Andreas Mazur, Henning Peters, André Artelt, Lukas Koller, Christoph Hartmann, Ansgar Trächtler and Barbara Hammer},
  booktitle    = {ICANN proceedings 2025}
  year         = {2025}
}

How to use the code

How to install this repository

This repository is not a standalone package, but rather a collection of scripts to run the experiments and generate the plots. Hence, you need to clone the repository first:

git clone https://github.com/andreasMazur/VisMeshSegmentation.git

Afterwards, you need to install the required packages. This can be done by creating a new conda environment and installing:

conda create -n plausible_cfs python=3.12
pip install -r requirements.txt

Running the experiments and generating the plots

The code is structured in the following way:

1. You need to run the training experiments first. This can be done by executing the script run_experiments.py. Depending on how you have specified your logging directory path, this will create a folder logs in which all the results are stored. Furthermore, you need to specify the path to this repository so that the experiments can find the dataset.
2. After the experiments are done, you can run the evaluation script run_evaluation.py to generate the plots. For this to work, you need to specify the path to the logging directory where the results are stored.