Hi all,

I'm performing a PCA on a matrix where the N columns with L length are the observations. I deduct the average column before computing the covariance matrix. The number of observations is much larger than the length of the observations. The covariance matrix is real and symmetric so I use the SelfAdjointEigenSolver.

The results seems fine, but it gives me L valid eigenvectors (non zero related eigenvalue). I think that a valid result should be L-1 valid eigenvectors.

Am I wrong?

Thanks in advance.

I've reduced the number of observations and compared with Matlab princomp() and the results are the same.

Problem solved.

Admins and mods: feel free to delete this thread due to is more a math question than an eigen question