The original code comes from this paper: "Learning Strategy-Aware Linear Classifiers" by Yiling Chen, Yang Liu, and Chara Podimata, which is accepted at NeurIPS2020. The paper is also publicly available here: https://arxiv.org/abs/1911.04004.
Folder cont_code/ contains the implementation for the continuous variant, and folder discr_code/ contains the implementation for the discretized variant.
It is run using python 2.7. and it has the following required packages:
*** cont_code/ ***
-- numpy
-- cvxpy
-- polytope
-- multiprocessing
-- sklearn
-- logging
-- os
*** extra packages required for discr_code/ ***
-- networkx
-- sys
File that controls the distribution of points: params.py.
File that controls the power of the agent (i.e., \delta): runner-dp.py
In order to run the code: python runner-dp.py.
Please see individual .py files for explanations regarding the functionality of each file.