This repository has code for replicating the experiments in High-Precision Model-Agnostic Explanations.

If you actually want to try anchors, use the code in this repository instead.

Inside a virtualenv, run:

pip install lime==0.1.1.29 sklearn matplotlib seaborn numpy argparse xgboost==0.4a30

Run:

python run_compute_explanations.py
python run_process_results.py
python make_graphs_and_table.py

Results will be in folder results