FANOK: Knockoffs in Linear Time
Abstract
We describe a series of algorithms that efficiently implement Gaussian model-X knockoffs to control the false discovery rate on large scale feature selection problems. Identifying the knockoff distribution requires solving a large scale semidefinite program for which we derive several efficient methods. One handles generic covariance matrices, has a complexity scaling as $O(p^3)$ where $p$ is the ambient dimension, while another assumes a rank $k$ factor model on the covariance matrix to reduce this complexity bound to $O(pk^2)$. We also derive efficient procedures to both estimate factor models and sample knockoff covariates with complexity linear in the dimension. We test our methods on problems with $p$ as large as $500,000$.
- Publication:
-
arXiv e-prints
- Pub Date:
- June 2020
- DOI:
- 10.48550/arXiv.2006.08790
- arXiv:
- arXiv:2006.08790
- Bibcode:
- 2020arXiv200608790A
- Keywords:
-
- Computer Science - Machine Learning;
- Statistics - Methodology;
- Statistics - Machine Learning
- E-Print:
- For code see https://github.com/qrebjock/fanok