README.md

Counterfactual Explanation for Time Series Data via Learned Saliency Maps

This is the repository for our paper titled "CELS: Counterfactual Explanation for Time Series Data via Learned Saliency Maps". This paper has been accepted at the 2023 IEEE International Conference on Big Data (Big Data).

Approach

Prerequisites and Instructions

All python packages needed are listed in pip-requirements.txt file and can be installed simply using the pip command.

Get the results for Coffee dataset by running

python3 main.py --pname CELS_Coffee --task_id 0 --run_mode turing --jobs_per_task 10 --samples_per_task 28 --dataset Coffee --algo cf --seed_value 1 --enable_lr_decay False --background_data train --background_data_perc 100 --enable_seed True --max_itr 1000 --run_id 0 --bbm dnn --enable_tvnorm True --enable_budget True --dataset_type test --l_budget_coeff 1 --run 1 --l_tv_norm_coeff 1 --l_max_coeff 1

The results would be saved into the bigdata_cels folder

Data

The data used in this project comes from the UCR archive.

Reference

If you re-use this work, please cite:

@inproceedings{li2023cels, title={Cels: Counterfactual explanations for time series data via learned saliency maps}, author={Li, Peiyu and Bahri, Omar and Boubrahimi, Souka{"\i}na Filali and Hamdi, Shah Muhammad}, booktitle={2023 IEEE International Conference on Big Data (BigData)}, pages={718--727}, year={2023}, organization={IEEE} }