Matlab eig command

[Pinaki]

Hi all,
I have found a problem with Malab R11 and R12 when I used the eig command.For symmetric square Matrix if I use the eig command to find an eigen value and eigen vector both versions work fine and give the same answer.But for a nonsymmetric square matrix matlab R11 and R12 are giving different results.So for certain algorithms for beamforming like robust beamformer are performing different way as they are highly sensitive regarding the choice of eigenvector.I have found matlab R11 is performing better than Matlab R12 in terms of the optimal performance of the robust beamformer.So I am wondering why Matlab R11 and R12 versions are choosing different eigenvectors for same nonsymmetric matrix.
Thanks,
Pinaki

 

[dileeprd]

you shouldnot worry if the different eigen vectors are linearly dependent.

i.e., matlabR11 could be giving the solution as [1 2 1]
and if matlabR12 gives as [0.5 1 0.5], they are still the same.

 

[VisiGoth]

I have found that it is better to augment the matrix and use qr in MATLAB for beam forming problems. Some people I know insist on using SVD, for better control over term culling. I thought only some unix systems had a noticable difference between R11 and R12.