Next: Intuitive Description of
Up: No Title
Previous: Convergence of Conjugate
- Note in eigenvalues: In Jacobi/SOR, ``large'' eigenvalues
converged slowly, ``small'' eigenvalues converged fast.
- But the matrices aren't the same.
- Conjugate gradient (CG) uses complete matrix A
- SOR uses decomposed matrix
-
eigenvalues determine SOR
eigenvalues determine CG
- Large eigenvalues of A
small eigenvalues of
SOR and vice versa.
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu