Re: Good eigenvalue/eigenvector test problems for eigs?

[Sebastien Loisel]

Yes, its supposed to be well behaved, but consider

octave:1> n=200;
octave:2> A = spdiags([ones(n,1),4*ones(n,1),-ones(n,1)],[-2,0,2],n,n);

Ok, that's not the same matrix I was talking about, but it's also well conditioned.

  -7.2806 + 0.0000i
  -7.1793 + 0.0000i
  -6.4207 + 0.0000i
   5.8174 + 0.0000i
  -5.6617 - 0.1134i
  -5.6617 + 0.1134i

That's really weird. I don't get that in Matlab. And it shouldn't even be a hard test case... Doesn't the QR iteration work just fine on this stuff?

Sébastien Loisel