Open
Description
We should add an early stop in QRCP.
Given a threshold parameter given by a user, perform QRCP until step k
where the norm of the trailing matrix (maybe the largest norm of all the norms of all the trailing updated vectors) is less than given threshold.
That way the cost of factorization is m * n * k
instead of m * n * n
.
For low rank m
-by-n
matrix A
, there is a significant advantage.
k
is the pseudo-rank of A
. An estimation of what the rank of A
could be.