ZSCDumin/GANITE

Codebase for GANITE: Estimation of Individualized Treatment Effects using GANs - ICLR 2018

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.

• Twin data: http://data.nber.org/data/linked-birth-infant-death-data-vital-statistics-data.html

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

Code explanation

(1) data_loading.py

• Transform raw twins data to preprocessed ITE data (X, T, Y, Potential Y)

(2) metrics.py (a) PEHE

• Precision in Estimation of Heterogeneous Effect (b) ATE
• Average Treatment Effect

(3) ganite.py

• Use observed features, treatments and factual outcomes to estimate the potential outcomes

(4) main_ganite.py

• Report PEHE and ATI for the twin dataset with GANITE

(5) utils.py

• Some utility functions for GANITE.

Command inputs:

• data_name: twin
• train_rate: the ratio of training data
• h_dim: hidden dimensions
• iterations: number of training iterations
• batch_size: the number of samples in each batch
• alpha: hyper-parameter to adjust the loss importance

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

• test_y_hat: estimated potential outcomes
• metric_results: PEHE and ATE