NUMERICAL LINEAR ALGEBRA

MAT/08 - 6 CFU - 2° semestre

Docente titolare dell'insegnamento

ARMANDO COCO
Email: armando.coco@unict.it
Edificio / Indirizzo: Dipartimento di Matematica e Informatica, Studio 310
Telefono: 095 7303010
Orario ricevimento: Lunedì e martedì dalle 11 alle 13.


Obiettivi formativi

Description

The module covers the design and analysis of numerical algorithms to solve or accurately approximate problems from linear algebra, such as linear systems and eigenvalue problems.

The module also aims at providing solid implementation skills by developing small software programs of the different numerical algorithms, with applications to real-world problems.



Contenuti del corso

Linear systems Reminders on some special matrices and their properties

Direct methods for full matrices linear systems: LU and QR factorization, Gaussian elimination, pivoting, Doolittle and Crout. Thomas algorithm. Preconditioning techniques: ILU, ILU(p), Incomplete Cholesky preconditioning.

Iterative methods for sparse matrices linear systems: Jacobi, Gauss-Seidel, SOR, SSOR, Krylov methods, Arnoldi orthogonalization, FOM and GMRES, Multigrid methods.

Eigenvalues

Direct methods

Non symmetric matrices: Power and inverse power methods. Similarity transformations: Householder and Givens. Simultaneous iteration, QR algorithm without and with shift.

Symmetric matrices: Tridiagonal QR iteration, Rayleigh ratio iteration, Divide & Conquer, Jacobi, Bisection, Sturm sequencies.

Iterative methods

Arnoldi method for non symmetric matrices and Lanczos method for symmetric matrices.




Apri in formato Pdf English version