Skip to content

Latest commit

 

History

History
135 lines (99 loc) · 3.51 KB

README.md

File metadata and controls

135 lines (99 loc) · 3.51 KB
Raw

Welcome to RISE!

Python package RISE for paper "RISE: Robust Individualized Decision Learning with Sensitive Variables", which has recently been accepted by NeurIPS 2022 for publication [paper].

TL;DR: We introduce RISE, a robust individualized decision learning framework to improve the worst-case outcomes of individuals caused by sensitive variables that are unavailable at the time of decision.

Keywords: causal inference, individualized treatment rules, sensitive variables, robustness.

To install package

git clone https://github.com/ellenxtan/RISE.git
cd RISE
pip install .

Dependencies

Tensorflow, Keras, sklearn, SHAP (SHapley Additive exPlanations), pandas, numpy, and all of their respective dependencies.

Examples to run RISE

With default parameters (no tuning)

from rise import utils as utils
from rise import rise as r
import pandas as pd

train_df = pd.read_csv("data/cont_df_train.csv")
test_df = pd.read_csv("data/cont_df_test.csv")

Y_name = "Y"
A_name = "A"
X_names = ["X1","X2","X3","X4","X5","X6"]
S_names = ["S"]
is_rct = False
is_class = False
s_type = "cont"
qua_use = 0.25

out, out_mod = r.rise(train_df, test_df, Y_name, A_name, X_names, S_names, 
                      is_rct, is_class, s_type, qua_use, 
                      is_tune=False, param_grid=None, is_plot=True, seed=12345)

Hyperparameter tuning

from rise import utils as utils
from rise import rise as r
import pandas as pd

train_df = pd.read_csv("data/cont_df_train.csv")
test_df = pd.read_csv("data/cont_df_test.csv")

Y_name = "Y"
A_name = "A"
X_names = ["X1","X2","X3","X4","X5","X6"]
S_names = ["S"]
is_rct = False
is_class = False
s_type = "cont"
qua_use = 0.25
is_tune = True
param_grid = dict(layers=[2,3], 
                  nodes=[256,512,1024], 
                  dropouts=[0.2,0.4], 
                  acts=["sigmoid","relu"], 
                  opts=["adam","nadam"], 
                  bsizes=[32,64,128], 
                  n_epochs=[50,100,200]
                  )

out, out_mod = r.rise(train_df, test_df, Y_name, A_name, X_names, S_names, 
                      is_rct, is_class, s_type, qua_use, 
                      is_tune, param_grid, is_plot=True, seed=12345)

Output

You may obtain the predicted decision rule, vulnerable index, and evaluation metrics including objective and value functions by calling:

out.A_pred
out.is_vul
out.obj_all
out.val_all
out.obj_vul
out.val_vul
out.model

SHAP plot for interpretation

Visualizations about feature importance in the final classification models to help interpret important covariates in making the decisions.

from rise import plots as p

p.plot_shap(out_mod, save_path="./")

fig_rise

Paper code

Code is under folder code_for_paper. Run script.py to get started.

  • script.py: starter code
  • loop.py: wrapper function for data analysis
  • imports.py: imported libraries
  • myfuns.py: utility functions
  • dat.py: functions for data

Cite

If you find this repository useful, please cite:

@inproceedings{
  tan2022rise,
  title={{RISE}: Robust Individualized Decision Learning with Sensitive Variables},
  author={Xiaoqing Tan and Zhengling Qi and Christopher Warren Seymour and Lu Tang},
  booktitle={Advances in Neural Information Processing Systems},
  editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
  year={2022},
  url={https://openreview.net/forum?id=-IHPcl1ZhF5}
}