You are on page 1of 2

CS 450 / CSE 401 / MATH 450 / ECE 491 Fall 2011

Numerical Analysis HW 4
Instructor: Anil N. Hirani University of Illinois at Urbana-Champaign

Due: 10/20/2011

Problem 1 [5pt] Let A= 1 4 . 1 1

Your answers to the following questions should be numeric and specic to this particular matrix, not just the general denitions. (a) [0.5pt] What is the characteristic polynomial of A? (b) [0.5pt] What are the roots of the characteristic polynomial of A? (c) [0.5pt] What are the eigenvalues of A? (d) [0.5pt] What are the eigenvectors of A? (e) [0.5pt] Perform on iteration of power iteration on A, using x0 = [1 1]T as a starting vector. (f) [0.5pt] To what eigenvector of A will power iteration ultimately converge? (g) [0.5pt] What eigenvalue estimate is given by the Rayleigh quotient, using the vector x = [1 1]T ? (h) [0.5pt] To what eigenvector of A would inverse iteration ultimately converge? (i) [0.5pt] What eigenvalue of A would be obtained if inverse iteration were used with shift = 2? (j) [0.5pt] If QR iteration were applied to A, to what form would it converge: diagonal or triangular? Why?

Problem 2 [2pt] Give an example of a 2 2 matrix A and a nonzero starting vector x0 such that power iteration fails to converge to the eigenvector corresponding to the dominant eigenvalue of A.

Problem 3 [2pt] Let A be an n n real matrix of rank one. (a) [0.5pt] Show that A = uv T for some nonzero real vectors u and v. (b) [0.5pt] Show that uT v is an eigenvalue of A.

(c) [0.5pt] What are the other eigenvalues of A? (d) [0.5pt] If power iteration is applied to A, how many iterations are required for it to converge exactly to the eigenvector corresponding to the dominant eigenvalue?

Problem 4 [1pt] (a) [0.5pt] What are the eigenvalues of the Householder transformation H =I 2 where v is any nonzero vector? (b) [0.5pt] What are the eigenvalues of the plane rotation G= where c2 + s2 = 1? c s s c , vv T , vT v

Problem 5 [1pt] Implement inverse iteration with a shift to compute the eigenvalue nearest to 2, and a corresponding normalized eigenvector, of the matrix 6 2 1 A = 2 3 1 . 1 1 1 You may use an arbitrary starting vector.

Problem 6 [1pt] Write a program implementing Rayleigh quotient iteration for computing an eigenvalue and corresponding eigenvector of a matrix. Test your program on the matrix in the previous exercise, using a random starting vector.

Problem 7 [2pt] Consider the generalized eigenvalue problem Kx = M x derived from the spring-mass system given in Example 4.1 and illustrated in Fig. 4.1. For purposes of this problem, assume the values k1 = k2 = k3 = 1 , m1 = 2, m2 = 3, and m3 = 4, in arbitrary units. For this particular problem, the mass matrix M is diagonal, so there is no harm in converting the generalized eigenvalue problem to a standard eigenvalue problem. Taking this approach, determine all three natural frequencies and modes of vibration for the system, using any combination you choose of the power and inverse iteration methods (you may use shifts, or deation, or both).

You might also like