Thread Subject: Eigen value and Eigen vector?

Hi,

Matlab function [V,D]=eig() would generate eigenvalues in
ascending order and the associated eigenvectors.

Is there an easy way to generate eigenvalues and
associated eigenvectors in descending order?

Thanks,
Stephen

On Aug 16, 4:20 pm, "stephen " <huan...@nationwide.com> wrote:
> Hi,
>
> Matlab function [V,D]=eig() would generate eigenvalues in
> ascending order and the associated eigenvectors.
>
> Is there an easy way to generate eigenvalues and
> associated eigenvectors in descending order?
>
> Thanks,
> Stephen

There's probably an easier way, but, if not, try:
 [V,D]=eig()
 V = V(:,length(V(:,1)):-1:1) ;
 D = D(length(D(1,:)):-1:1,length(D(:,1)):-1:1) ;

Joao

"J.N." <joao.natali@gmail.com> schrieb im Newsbeitrag
news:1187310068.805535.110740@19g2000hsx.googlegroups.com...
> On Aug 16, 4:20 pm, "stephen " <huan...@nationwide.com> wrote:
>> Hi,
>>
>> Matlab function [V,D]=eig() would generate eigenvalues in
>> ascending order and the associated eigenvectors.
>>
>> Is there an easy way to generate eigenvalues and
>> associated eigenvectors in descending order?
>>
>> Thanks,
>> Stephen
>
> There's probably an easier way, but, if not, try:
> [V,D]=eig()
> V = V(:,length(V(:,1)):-1:1) ;
> D = D(length(D(1,:)):-1:1,length(D(:,1)):-1:1) ;
>
> Joao
>

Hi,
slightly easier: use "end" instead of length:
V = V(:, end:-1:1);
D = D(end:-1:1,end:-1:1)

Titus

Titus:
<SNIP the big easy...

> > D = D(length(D(1,:)):-1:1,length(D(:,1)):-1:1) ;

> slightly easier: use "end" instead of length:
> D = D(end:-1:1,end:-1:1)

... or, yet another solution

     D=rot90(D,2)

us

Thanks for all the responses, guys.

I am a little confused now as I look at the results. I
have an Excel VBA program which could calculate the
Eigenvalues in descending order and the associated
Eigenvetors. If I use the suggested method to calculate
the same things in Matlab, some values of Eigenvector are
exactly the same while some values of Eigenvector are the
same in absolute terms but have opposite signs (plus vs
minus). I use formula V*D*V’ to check if they are getting
the same correlation matrix, and they are.

Does anyone have any idea what’s going on here?

Thanks,
Stephen




"us " <us@neurol.unizh.ch> wrote in message
<fa3kac$rlr$1@fred.mathworks.com>...
> Titus:
> <SNIP the big easy...
>
> > > D = D(length(D(1,:)):-1:1,length(D(:,1)):-1:1) ;
>
> > slightly easier: use "end" instead of length:
> > D = D(end:-1:1,end:-1:1)
>
> ... or, yet another solution
>
> D=rot90(D,2)
>
> us

"stephen " <huangj5@nationwide.com> schrieb im Newsbeitrag
news:fa4a5e$35c$1@fred.mathworks.com...
> Thanks for all the responses, guys.
>
> I am a little confused now as I look at the results. I
> have an Excel VBA program which could calculate the
> Eigenvalues in descending order and the associated
> Eigenvetors. If I use the suggested method to calculate
> the same things in Matlab, some values of Eigenvector are
> exactly the same while some values of Eigenvector are the
> same in absolute terms but have opposite signs (plus vs
> minus). I use formula V*D*V’ to check if they are getting
> the same correlation matrix, and they are.
>
> Does anyone have any idea what’s going on here?
>
> Thanks,
> Stephen
>
>
>
>
> "us " <us@neurol.unizh.ch> wrote in message
> <fa3kac$rlr$1@fred.mathworks.com>...
>> Titus:
>> <SNIP the big easy...
>>
>> > > D = D(length(D(1,:)):-1:1,length(D(:,1)):-1:1) ;
>>
>> > slightly easier: use "end" instead of length:
>> > D = D(end:-1:1,end:-1:1)
>>
>> ... or, yet another solution
>>
>> D=rot90(D,2)
>>
>> us
>

Hi Stephen,
yes I have ;-)
Eigenvectors are not unique in the following sense: if x is an eigenvector,
k*x for every k~=0 is an eigenvector as well, since an eigenvector is a non
trivial vector that solves
A * x = lambda * x
so: for any scalar k~=0
A * (k*x) = k * (A * x) = k * (lambda * x) = lambda * (k * x),
so k*x is an eigenvector as well, especially for k=-1...

Titus

On Aug 16, 4:20 pm, "stephen " <huan...@nationwide.com> wrote:
> Hi,
>
> Matlab function [V,D]=eig() would generate eigenvalues in
> ascending order and the associated eigenvectors.
>
> Is there an easy way to generate eigenvalues and
> associated eigenvectors in descending order?

1. Use EIGS instead of EIG
2. Use FLIPLR and FLIPUD

Hope this helps.

Greg