AI4LIFE-GROUP / ROAR

ROAR

Requirements

To run this code first install the packages in requirements.txt. This code also requires custom extensions of the open sourced actionable-recourse and LIME libraries contained in actionable-recourse/ and lime/ respectively. Follow the requirements instructions and run python setup.py in each folder.

Organization

data.py

This file contains the data loading and pre-processing modules.

model.py

This file contains the LR, DNN, and SVM implementations.

recourse_methods.py

This file contains the implementations of all the recourse methods. Our method, ROAR, is called RobustRecourse, CFE is called counterfactual_recourse, AR is called actionable_recourse, and CR is called CausalRecourse.

recourse_utils.py

This file contains helper functions for the recourse methods and evaluation.

run_rw.py

This file generates recourses for our real world data experiments.

run_sim.py

This file generates recourses for our synthetic experiments.

run_delta.py

This file generates recourses for our experiments varying $\delta_max$ (see Appendix).