===firstname:
Kathryn
===firstname3:
Marcel
===affil6:

===lastname3:
Schweitzer
===email:
lund-nguyen@temple.edu
===keyword_other2:

===lastname6:

===affil5:

===lastname4:
Szyld
===lastname7:

===affil7:

===postal:
Department of Mathematics, Temple University
1805 N. Broad Street
Rm 638 Wachman Hall
Philadelphia, PA 19122
===ABSTRACT:
We present a new iterative method for computing $f(A) \vb$, derived from a relationship between the standard Lanczos method and a Gauss-Radau quadrature rule.  We show that this method, called the Radau-Lanczos method, converges when $A$ is Hermitian positive definite and $f$ is a Stieltjes function.  We also show that the restarted version of this method converges and present numerical results showing this method performing better than the standard Lanczos method in terms of attainable error norm and iteration count.
===affil3:
Bergische Universit\"{a}t Wuppertal
===title:
The Radau-Lanczos method for matrix functions
===affil2:
Bergische Universit\"{a}t Wuppertal
===lastname2:
Frommer
===firstname4:
Daniel B.
===keyword1:
APP_OTHER
===workshop:
no
===lastname:
Lund-Nguyen
===firstname5:

===keyword2:
Quantum chemistry/physics
===otherauths:

===affil4:
Temple University
===competition:
yes
===firstname7:

===firstname6:

===keyword_other1:
Iterative solvers for matrix functions
===lastname5:

===affilother:

===firstname2:
Andreas