11.07.2015 Views

Introduction to Scientific Computing Gauss-Seidel Method. Power ...

Introduction to Scientific Computing Gauss-Seidel Method. Power ...

Introduction to Scientific Computing Gauss-Seidel Method. Power ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Institute of <strong>Scientific</strong> <strong>Computing</strong>Technische Universität BraunschweigProf. Hermann G. Matthies, PhDDr. Alexander LitvinenkoWinter Term 2009-2010January 14, 2010<strong>Introduction</strong> <strong>to</strong> <strong>Scientific</strong> <strong>Computing</strong><strong>Gauss</strong>-<strong>Seidel</strong> <strong>Method</strong>. <strong>Power</strong> method.Assignment 11Exercise 1: Linear two-step methodConsider the linear two-step method(12 points)x n+2 + 4x n+1 − 5x n = h · (4f n+1 + 2f n ) (1)(a) Show that this method is consistent. Which order of consistency has the method ? (4 points)(b)Implement this method and solve the ODE˙u = uin the interval [0,1] with different step sizes h.u(0) = 1(2)(4 points)(c) Explain the numerical results (convergence). (4 points)Exercise 2: <strong>Gauss</strong>-<strong>Seidel</strong> <strong>Method</strong>Depending on the real valued parameter α we have a matrix A defined as( )1 αA =−α 2(12 points)and b = (b 1 ,b 2 ) ⊤ an arbitrary (but given) vec<strong>to</strong>r. For solving the system of equations we use the<strong>Gauss</strong>-<strong>Seidel</strong> method.(a) For which values of α does this method converge in general (using an arbitrary starting vec<strong>to</strong>r)and for which values does it diverge in general?(6 points)(b) Let b = (1,2) ⊤ . Solve the system of equations Ax = b numerically using the <strong>Gauss</strong>-<strong>Seidel</strong>method for α = − 3 2 + β100, β = 0,1,... ,300. Plot the numerical error and the number of iterationsvs. the different values of alpha.(6 points)Exercise 3: <strong>Power</strong> method (power iteration)(14 points)Sometimes it is very important <strong>to</strong> compute only the largest eigenvalue of a large matrix (withoutcomputing all other eigenvalues). The power method (or power iteration) performs this task.(a) Implement this method as a MATLAB procedure. (6 points)(b)Generate a 100*100 matrix in MATLAB and use your procedure <strong>to</strong> compute the largest eigenvalue.Use also MATLAB “eigs” procedure <strong>to</strong> check yourself. Pay attention that the power method


works not for all matrices.(6 points)(c) Compare computing times. (2 points)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!