README.md

In davidaknowles/implicit: Implicitly scaled matrices and products of two matrices for use in diagonal CCA

implicit

This is an R package to allow performing PCA and diagonal CCA with implicitly scaled matrices. For example, X (dimension N x P1) and Y (dimension N x P2) can be sparse matrices and we would like to perform PCA on scale(X) or (diagonal) CCA on scale(X) and scale(Y). The challenge is that scale(X) is no longer sparse. implicit implements scaled_matrix and centered_matrix which implicitly scale and/or center an underlying matrix and allows left and right multiplication with a vector (the key operation for fast SVD using irlba). This avoids ever creating large dense matrices but gives the same results up to (approximately) machine precision.

davidaknowles/implicit documentation built on Dec. 19, 2021, 9:06 p.m.