I am using Arpack to compute the eigenvalues of the problem \lambda Mx = Ax with reverse shift method with complex shift. Then, I use znaupd e zneupd. Everything seems to be ok. I wrote a function that takes the eigenvalues and the eigenvectors computed with Arpack and put them in the original problem: \lambda Mx -Ax = residual. The norm of the residual is supposed to be almost zero or very small. Anyway, I obtain an error of 10^-3 that is quite high! Do you have any idea where the problem can come from?
Asked
Active
Viewed 316 times
2
-
Maybe, it is more appropriate to ask at scicomp.stackexchange.com See this: http://scicomp.stackexchange.com/questions/10222/generalized-eigenvalue-problem-using-arpack The documentation advises to convert problem into standard form – John_West Feb 06 '15 at 10:23
-
Thank you, but my matrix M is real and symmetric... I'll try to ask to scicomp.stackexchange.com – Britomarti Feb 06 '15 at 10:43