MTH 655 and MTH 659 (Numerical Analysis)
Large scale scientific computing methods - Fall 2009
Links
General information
Textbook and resources
Labs
Labs/schedule

Date Schedule/assignment
Mon 1/7
  1. Administration: fill out questionnaire.html
  2. Overview - slides in Blackboard.
Wed 1/09 Functional iteration and Newton's method in R: convergence. Read Chapter 5 from [Kelley]. and preliminaries, if necessary.
Fri 1/11 LAB 1 (MLC Kidd 108J) computer lab. You can come as early as 8:30.
Mon 1/14 Implementation issues for Newton's method (improve primitive algorithm). A-posteriori estimates for error.
Wed 1/16 Newton's in N-dimensions. Newton-like methods: chord, secant, Broyden, inexact Newton's methods.
Fri 1/18 LAB 2 (MLC Kidd 108J) computer lab. You can use the primitive template newton2d.m
Wed 1/23 Relaxation algorithms for nonlinear problems. Newton for semismooth functions. Application: solving problems under inequality constraints.
Fri 1/25 LAB 3 (MLC Kidd 108J) computer lab.
  • Linux for beginners and C-shell tutorials
  • Fortran tutorial and Fortran reference
  • For C tutorials - well, just google-it up
  • Please read about ssh and scp .
  • You must be able to use some text editor under Unix. I suggest emacs or vi. Please look up the basics. Or ask me.
  • Mon 1/28 More on solving problems under inequality constraints: semismooth Newton methods.
    Wed 1/30 Jacobi, Gauss-Seidel and theor relaxation versions. Conditions on convergence involving diagonal dominance and spectral radius. Gershgorin circle theorem.
    Fri 2/1 Continue LAB 3 in (MLC Kidd 108J) computer lab.
    Mon 2/4 Spectral properties of the "discrete laplacian". Spectrum of weighted Jacobi iteration matrix.
    Wed 2/6 Smoothing properties of weighted Jacobi method. Idea of coarse grid correction.
    Fri 2/8 LAB 4 (MLC Kidd 108J computer lab). Exploration of Newton and semi-smooth Newton methods. Stationary and relaxation iterative solvers.
    You can use codes templates semicon.m and semilinear.m.
    Mon 2/11 Coarse grid correction wrap-up. Full multigrid idea and performance.
    Wed 2/13 Non-stationary iterative methods: projectiob methods and Krylov subspaces.
    Fri 2/15 LAB 5 (MLC Kidd 108J computer lab): inexact Newton-Krylov methods
    Mon 2/18 Non-stationary iterative methods and Krylov spaces; steepeest descent (SD), conjugate gradients (CG), and GMRES.
    Wed 2/20 Rosenbrock banana. Preconditioning: algebraically or yusing physics based preconditioners.
    Fri 2/22 LAB 6 (MLC Kidd 108J computer lab): preconditioning in nonstationary methods.
    Mon 2/25 Basics of parallel computing with mypi.f example.
    Presentation 655-jacobi.pdf .
    Wed 2/27 Overview of MPI (Message Passing Interface).
    Reduce and send-receive operations. Scalability in parallel computing.
    3/1 LAB7 (MLC Kidd 108J computer lab): working on the cluster.
    Mon 3/4 Domain decomposition methods: overlapping: Schwarz additive method. Block Jacobi.
    Wed 3/6 Domain decomposition methods: nonoverlapping; D-N and N-N methods.
    3/8 (MLC Kidd 108J computer lab): finishing all assignments
    Mon 3/11 Monte Carlo methods. (LABS 1-4 must be turned in).
    Wed 3/13 Introduction to statistical mechanics: molecular dynamics and lattice methods. (ALL LABS must be turned in).
    3/8 LAB REVIEW