Toward Restarting Strategies Tuning for a Krylov Eigenvalue Solver
نویسندگان
چکیده
Krylov eigensolvers are used in many scientific fields, such as nuclear physics, page ranking, oil and gas exploration, etc... In this paper, we focus on the ERAM Krylov eigensolver whose convergence is strongly correlated to the Krylov subspace size and the restarting vector v0, a unit norm vector. We focus on computing the restarting vector v0 to accelerate the ERAM convergence. First, we study different restarting strategies and compare their efficiency. Then, we mix these restarting strategies and show the considerable ERAM convergence improvement. Mixing the restarting strategies optimizes the ”numerical efficiency” versus ”execution time” ratio as we do not introduce neither additionnal computation nor
منابع مشابه
A Hamiltonian Krylov-Schur-type method based on the symplectic Lanczos process
We discuss a Krylov-Schur like restarting technique applied within the symplectic Lanczos algorithm for the Hamiltonian eigenvalue problem. This allows to easily implement a purging and locking strategy in order to improve the convergence properties of the symplectic Lanczos algorithm. The Krylov-Schur-like restarting is based on the SR algorithm. Some ingredients of the latter need to be adapt...
متن کاملSome new restart vectors for explicitly restarted Arnoldi method
The explicitly restarted Arnoldi method (ERAM) can be used to find some eigenvalues of large and sparse matrices. However, it has been shown that even this method may fail to converge. In this paper, we present two new methods to accelerate the convergence of ERAM algorithm. In these methods, we apply two strategies for the updated initial vector in each restart cycles. The implementation of th...
متن کاملConvergence of Polynomial Restart Krylov Methods for Eigenvalue Computations
Krylov subspace methods have proved effective for many non-Hermitian eigenvalue problems, yet the analysis of such algorithms is involved. Convergence can be characterized by the angle the approximating subspace forms with a desired invariant subspace, resulting in a geometric framework that is robust to eigenvalue ill-conditioning. This paper describes a new bound on this angle that handles th...
متن کاملEigenvalue Solvers for Modeling Nuclear Reactors on Leadership Class Machines
Three complementary methods have been implemented in the code Denovo that accelerate neutral particle transport calculations with methods that use leadership-class computers fully and effectively: a multigroup block (MG) Krylov solver, a Rayleigh Quotient Iteration (RQI) eigenvalue solver, and a multigrid in energy (MGE) preconditioner. The MG Krylov solver converges more quickly than Gauss Sei...
متن کاملConvergence of Restarted Krylov Subspaces to Invariant Subspaces
The performance of Krylov subspace eigenvalue algorithms for large matrices can be measured by the angle between a desired invariant subspace and the Krylov subspace. We develop general bounds for this convergence that include the effects of polynomial restarting and impose no restrictions concerning the diagonalizability of the matrix or its degree of nonnormality. Associated with a desired se...
متن کامل