Home

Awesome

Codebase for "GANITE"

Authors: Jinsung Yoon, James Jordon, Mihaela van der Schaar

Reference: Jinsung Yoon, James Jordon, Mihaela van der Schaar, "GANITE: Estimation of Individualized Treatment Effects using Generative Adversarial Nets", International Conference on Learning Representations (ICLR), 2018.

Paper link: https://openreview.net/forum?id=ByKWUeWA-

Contact: jsyoon0823@gmail.com

This directory contains implementations of GANITE framework for individualized treatment effect estimations using a real-world dataset.

To run the pipeline for training and evaluation on GANITE framwork, simply run python3 -m main_ganite.py.

Code explanation

(1) data_loading.py

(2) metrics.py (a) PEHE

(3) ganite.py

(4) main_ganite.py

(5) utils.py

Command inputs:

Note that network parameters should be optimized.

Example command

$ python3 main_ganite.py --data_name twin --train_rate 0.8 
--h_dim 30 --iteration 10000 --batch_size 256 --alpha 1 

Outputs