Deep Knockoffs

Approximate knockoffs for model-free variable selection.

DeepKnockoffs is a software package for sampling approximate model-X knockoffs using deep generative models.

The methods described in the paper below are implemented in Python with the help of the PyTorch library.

The code is publicly available from GitHub: https://github.com/msesia/deepknockoffs.

Reference

“Deep Knockoffs”,
Yaniv Romano, Matteo Sesia and Emmanuel Candès. arXiv:1811.06687 (2018). Link to the paper.

License: GPLv3

Main features

Generation of approximate knockoff copies.
Goodness-of-fit diagnostics for knockoffs.
Knockoff filter for variable selection.

Authors

Matteo Sesia and Yaniv Romano.