**Unformatted text preview: **to N (I − A) along R (I − A). Below is a formal summary
of our observations concerning Ces`ro summability.
a Copyright c 2000 SIAM Buy online from SIAM
http://www.ec-securehost.com/SIAM/ot71.html Buy from AMAZON.com
7.10 Diﬀerence Equations, Limits, and Summability
http://www.amazon.com/exec/obidos/ASIN/0898714540 633 Ces` ro Summability
a
A ∈ C n×n is Ces`ro summable if and only if ρ(A) < 1 or else
a
ρ(A) = 1 with each eigenvalue on the unit circle being semisimple. • It is illegal to print, duplicate, or distribute this material
Please report violations to [email protected] • When it exists, the Ces`ro limit
a
I + A + · · · + Ak−1
=G
k→∞
k
lim D
E (7.10.36) is the projector onto N (I − A) along R (I − A). T
H • G = 0 if and only if 1 ∈ σ (A) , in which case G is the spectral
projector associated with λ = 1. • If A is convergent to G, then A is summable to G, but not
conversely. IG
R Since the projector G onto N (I − A) along R (I − A) plays a prominent
role, let’s consider how G might be computed. Of course, we could just iterate
on Ak or (I + A + · · · + Ak−1 )/k, but this is ineﬃcient and, depending on the
proximity of the eigenvalues relative to the unit circle, convergence can be slow—
averaging in particular can be extremely slow. The Jordan form is the basis for
the theoretical development, but using it to compute G would be silly (see
p. 592). The formula for a projector given in (5.9.12) on p. 386 is a possibility,
but using a full-rank factorization of I − A is an attractive alternative.
A full-rank factorization of a matrix Mm×n of rank r is a factorization O
C Y
P M = Bm×r Cr×n , where rank (B) = rank (C) = r = rank (M). (7.10.37) All of the standard reduction techniques produce full-rank factorizations. For
example, Gaussian elimination can be used because if B is the matrix of basic
columns of M, and if C is the matrix containing the nonzero rows in the
reduced row echelon form EM , then M = BC is a full-rank factori...

View
Full Document