Skip to content
Design of Simulations using WGAN
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data/original_data
docs
trained_models
.gitignore
README.md
requirements.txt
setup.py
wgan.py

README.md

Guide to Simulating Economic Datasets using WGANs

This repository contains wgan, a python module built on PyTorch for using WGANs to simulate from joint and conditional distributions of economic datasets. Brief documentation and the API for the module is available here.

This Google Colab Notebook contains a tutorial for the code, including how to install requirements, and estimate and simulate from the resulting models, using a free Google GPU. The tutorial simulates from the Lalonde-Dehejia-Wahba dataset, as described in detail in the following paper.

Athey, Susan, Guido Imbens, Jonas Metzger, and Evan Munro. "Using Wasserstein Generative Adversial Networks for the Design of Monte Carlo Simulations." arXiv:1909.02210. September 2019.

The data folder contains the raw Lalonde-Dehejia-Wahba data.

Installation Instructions

For running the WGAN code outside of a Google Colab environment an installation of python3 is required.

In order to install this package dependencies, please use requirements.txt:

git clone https://www.github.com/gsbDBI/ds-wgan
cd ds-wgan
pip install -r requirements.txt 

In addition, in order to install the package itself, you can run:

python setup.py develop
You can’t perform that action at this time.