An algorithm for rank estimation and subspace tracking
Abstract
This article presents an URV-based matrix decomposition, the truncated URV decomposition, and an updating algorithm for it. The complexity of the updating is [image omitted] for an m-by- n matrix of rank r. The theoretical and numerical results presented shows that the decomposition can be a good alternative to the singular value decomposition.