Simulations

All code necessary to reproduce the simulation results from our paper

@inproceedings{
    goebler2025gresit,
    title={Nonlinear Causal Discovery for Grouped Data},
    author={Goebler, K., Windisch, T., Drton, M.},
    booktitle={The 41st Conference on Uncertainty in Artificial Intelligence},
    year={2025},
}

is contained in the folder simulation.

Essentially, executing ./run_simulations.sh executes the benchmarks for different numbers of nodes, groups, and samples, sequentially. In our experiments, we use slurm to massively parallelize the individual runs.