National Instruments 370757C-01 用户手册
Chapter 2
Robustness Analysis
2-16
ni.com
VOPT=ssv(M,{scaling="OPT"})
VOPT (a scalar) = 2.43952
VSVD = max(svd(M))
VSVD (a scalar) = 2.65886
osscale( )
[v, vD] = osscale(M)
The
osscale( )
function scales a matrix using the Osborne Algorithm.
A diagonal scaling D
OS
is found that minimizes the Frobenius norm of
, which is the square root of the sum of the squares of its
singular values. If M is reducible,
osscale( )
may encounter a divide
by zero. To avoid this, use
ssv( )
with the Osborne scaling option:
[v,vD]=ssv(M,{scaling="OS"})
pfscale( )
[v, vD] = pfscale(M)
The
pfscale( )
function scales a matrix using the Perron-Frobenius
Algorithm. This scaling is optimal for matrices with all positive entries.
The matrix M must be irreducible for this function. If M is reducible,
use
The matrix M must be irreducible for this function. If M is reducible,
use
ssv( )
with the Perron-Frobenius scaling option instead:
[v,vD]=ssv(M,{scaling="PF"})
The optimum diagonal scaling is found for M using the Perron-Frobenius
theory of non-negative matrices. This scaling is given by
theory of non-negative matrices. This scaling is given by
where p and q are right and left eigenvectors of | associated with its largest
eigenvalue:
eigenvalue:
D
OS
MD
OS
1
–
D
i
PF
p
i
q
i
----
=
Mp
λ
max
p,
=
M
T
q
λ
max
q
,
=
p 0 q
≠ ≠