Repository containing the code for the experiments of TracInAD : Measuring Influence for Anomaly Detection accepted in the proceedings of IJCNN 2022.
To run experiments, datasets must be downloaded in folder data:
cd tracinad_wcci2022
bash ./data/data.sh
data.sh requires wget to be installed. For Mac user, replace with curl in data.sh.
To run the experiments, use the requirements.txt file to install the dependencies.
Using virtualenv:
cd TracInAD
virtualenv tracinad_env
source ./tracinad_env/bin/activate
pip install -r requirements.txt
Using conda:
cd TracInAD
conda create -n tracinad_env
conda activate tracinad_env
conda install --file requirements.txt
To run experiments:
cd TracInAD
source ./tracinad_env/bin/activate
bash run.sh -d arrhythmia
For other datasets, replace arrhythmia by a dataset contained in [thyroid, arrhythmia, kdd, kddrev].
If you use this code, please cite us:
@inproceedings{Thimonier_2022,
title={Trac{I}n{AD}: Measuring Influence for Anomaly Detection},
url={http://dx.doi.org/10.1109/IJCNN55064.2022.9892058},
DOI={10.1109/ijcnn55064.2022.9892058},
booktitle={2022 International Joint Conference on Neural Networks (IJCNN)},
publisher={IEEE},
author={Thimonier, Hugo and Popineau, Fabrice and Rimmel, Arpad and Doan, Bich-Lien and Daniel, Fabrice},
year={2022},
month=jul }