Skip to content
/ MCSECE Public

Multi-Criteria Approach for Selecting an Explanation From the Set of Counterfactuals Produced by an Ensemble of Explainers

License

Notifications You must be signed in to change notification settings

istepka/MCSECE

Repository files navigation

Multi-criteria approach for selecting an explanation from the set of counterfactuals produced by an ensemble of explainers

This is a repository for the paper: "Multi-criteria approach for selecting an explanation from the set of counterfactuals produced by an ensemble of explainers".

The paper is available at: https://arxiv.org/abs/2403.13940

1. Installation

git clone --recurse-submodules https://github.com/istepka/MCSECE
conda create -n <env_name> python==3.10
pip install -r requirements.txt

Go to ./modules and check if CARLA and CFEC modules are loaded there. If not get them using git submodule command.

2. Setup

  1. (compulsory) Unzip data/datasets.zip file into data/ folder.
  2. (optional - reproducibility) Unzip experiments/experiments_results.zip file into experiments/ folder.

3. Reproducing paper experiments

To reproduce paper experiments (tested on Ubuntu 20) from scratch follow the steps:

  1. Set your conda environment path in run_reproducibility.sh script.

  2. Generate explanations for the test datasets

    > chmod u+x run_reproducibility.sh
    > ./run_reproducibility.sh

    Now under ./experiments/data you should see a folder with your current date.
    This might take some time. In order to speed things up you can easily run this script in parallel for each dataset.

  3. (optional) After data is properly generated or after unziping data from original paper experiments/experiments_results.zip file you can run the following scripts to obtain visualizations and statistics:
    For Tables 2,3,4,5 (scores) and Figure 2 (barycentric plots)

    python3 experiments/experiment1.py

    For Table 6 and 7 (statistics)

    python3 experiments/experiment1_stats.py

    For radar plots

    python3 experiments/radar_graphs.py

    For pareto front plots

    python3 experiments/pareto_front_graphs.py

Cite as

@article{stepka2024multi,
  title={A Multi--Criteria Approach for Selecting an Explanation from the Set of Counterfactuals Produced by an Ensemble of Explainers},
  author={Stepka, Ignacy and Lango, Mateusz and Stefanowski, Jerzy},
  journal={International Journal of Applied Mathematics and Computer Science},
  volume={34},
  number={1},
  pages={119--133},
  year={2024}
}

About

Multi-Criteria Approach for Selecting an Explanation From the Set of Counterfactuals Produced by an Ensemble of Explainers

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published