National Instruments 370757C-01 用户手册

VOPT=ssv(M,{scaling="OPT"})

VSVD = max(svd(M))

[v, vD] = osscale(M)

The 

osscale( )

 function scales a matrix using the Osborne Algorithm. 

A diagonal scaling D

 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"})

[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

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

where p and q are right and left eigenvectors of | associated with its largest 
eigenvalue:

1

–

----

=

λ

p,

=

λ

,

=

p 0 q

≠ ≠