Can linear algebra create perfect knockoffs?

Abstract

As new Model-X knockoff construction techniques are developed, primarily concerned with determining the correct conditional distribution from which to sample, we focus less on deriving the correct multivariate distribution and instead ask if ``perfect'' knockoffs can be constructed using linear algebra. Using mean absolute correlation between knockoffs and features as a measure of quality, we do produce knockoffs that are pseudo-perfect, however, the optimization algorithm is computationally very expensive. We outline a series of methods to significantly reduce the computation time of the algorithm.