Prev: automatically write m files from script
Next: Creating a vector from sub vector
From: d on 23 May 2010 16:49
"Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message
....
> > Something like this perhaps:
> >
> > [V,D]=eig(myMatrix);
> > [trash,idx]=max(abs(diag(D)));
> > PerronVector=V(:,idx);
>
> Not quite, first we have to note that using max(diag(D)) is correct. But even with that change, I don't believe there is any warranty the above would return Perron vector.
....
> for example. Both of course are not Perron's vector. That's where the difficulty arises.
>
> Bruno

thank you for responding.

I would like to add that I only need the vector corresponding to the largest (absolute) eigenvalue, therefore I would rather not compute all eigenvalues+eigenvectors. I'm dealing with quite large, somewhat sparse, matrices (~700x700).

D
From: Bruno Luong on 23 May 2010 17:05
"d " <dagkatan(a)yahoo.com> wrote in message <htc4bv$so6$1(a)fred.mathworks.com>...
>
>
> I would like to add that I only need the vector corresponding to the largest (absolute) eigenvalue, therefore I would rather not compute all eigenvalues+eigenvectors. I'm dealing with quite large, somewhat sparse, matrices (~700x700).
>

Do you have an example where

[V D]=eigs(A,1,'lm')

fails?

Bruno
From: d on 23 May 2010 17:06
"Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <htc3p7$l8f$1(a)fred.mathworks.com>...
> I suggest to use LINPROG (opt toolbox) to eventually "fix" the eigen vector.
>
> [V D]=eig(M);
> lambda = diag(D);
..
..
..
> x = linprog(dummy, A, b, Aeq, beq);
> Vperron = V*x
>
> % Bruno


hi Bruno, just saw your suggestion now.
it actually had no problem with the size of the matrix, and was even faster then another code I found:
http://www.mathworks.de/matlabcentral/fileexchange/22763-perron-root-computation

the two methods gave similar results, which I think is a good sign.
thank you very much

D
From: Bruno Luong on 23 May 2010 17:17
"d " <dagkatan(a)yahoo.com> wrote in message <htc5bs$3ln$1(a)fred.mathworks.com>...

>
> hi Bruno, just saw your suggestion now.
> it actually had no problem with the size of the matrix, and was even faster then another code I found:
> http://www.mathworks.de/matlabcentral/fileexchange/22763-perron-root-computation
>
> the two methods gave similar results, which I think is a good sign.
> thank you very much

It looks like this code is from people who know about this kind of problem, so you should perhaps use their.

Bruno
First  |  Prev  | 
Pages: 1 2
Prev: automatically write m files from script
Next: Creating a vector from sub vector