Hi there, I Juste try to modify the SelfAdjointEigenSolver.classs: For users which have the input real symmetric matrix and only neeed compute first LARGEST k eigenvectors and eigenvalues

https://github.com/lukaspistelak/SelfAd ... .gitignore

I think this is very usefull function, only compute few wirst eigenvalues and corespond eigenvectors:

1. I just modify constructor and add param: int k:

SelfAdjointEigenSolver(const MatrixType& matrix,int ispower,int k)

2 I add method computePower(matrix, ispower,k);

Thaths all

(NOTE matrix vector product can be optim with .eval() and or .SelfADjointView()...

It seems toworks but after couple of days it sometimes return NAN so there is some programming mistake.....

So my question is that, IF Eigen Developers count with this in the next update? THanks FOR HELP!

Yes, at some point simple power iterations should be part of Eigen, but not as a method of SeflAdjointEigenSolver. It should be implemented in a separated class supporting any matrix types (dense, sparse, band, etc.). See a start there: https://bitbucket.org/ensigll/eigen/src ... at=default