[Trilinos-Users] [EXTERNAL] compute eigenvalues of an ill-conditioned matrix

[Klinvex, Alicia Marie]

Hello Da,

Can you tell us more about your problem?  For instance, is your graph undirected, i.e. is your matrix symmetric?  Are you looking for the largest eigenpairs or the smallest?  How many are you computing?  How accurate do they need to be?  Are you explicitly forming the matrix whose eigenvalues are being computed?  How high is a very high condition number?

Best wishes,
Alicia

-----Original Message-----
From: Trilinos-Users [mailto:trilinos-users-bounces at trilinos.org] On Behalf Of Da Zheng
Sent: Saturday, March 11, 2017 10:26 AM
To: trilinos-users at trilinos.org
Cc: Carey E Priebe <cep at jhu.edu>
Subject: [EXTERNAL] [Trilinos-Users] compute eigenvalues of an ill-conditioned matrix

Hello,

I'm trying to use the Anasazi eigensolvers to compute spectral embedding of a very large graph (|V| ~ 4 billion and |E| ~ 129 billion). I'm just curious if the eigensolvers such as Krylov Schur can compute eigenvalues with a high precision even if the sparse matrix has a very high condition number?

Thanks,
Da
_______________________________________________
Trilinos-Users mailing list
Trilinos-Users at trilinos.org
https://trilinos.org/mailman/listinfo/trilinos-users