Homework Questions: Numerical Methods
Homework Questions: Numerical Methods
Spring 2023
Problem Set 4 (Linear Algebraic Equations)
29. Calcualte the condition numbers cond(Ap) , p = 1,2, ∞ for the matrix
100 99
A = 99 98
(a) Use the algorithm for tridiagonal systems to decompose A = LU and solve Ax = b.
-14 -14 -10 00 00 10
A = 0 -1 4 -1 0 ;
b= 0 .
0 0 -1 4 -1 0
0 0 0 -1 4 0
(b) Solve the system Ax = b with A and b as in (a) using Gauss-Seidel iteration. Use the initial
guess xo = 0.
31. Consider the block tridiagonal system Ax = b. Suppose that each block of A is 2 × 2.
(a) Discuss how you can modify the Thomas algorithm to solve this system.
(b) Generalize to the case where each block is k × k.
(c) What is the OC in this case?
32. Consider the following method for producing the LU factorization of A. We write
A* d
A= T
c α
where A* is a square matrix of order (n-1), c and d are (n-1)×1vectors and α is a scalar. Factor A
as
L* 0 U* q
A= T .
m 1 0 γ
This method can be applied to the original matrix factoring each principal submatrix in the upper
left corner, in increasing order. Determine the formulas for q, m and γ given c, d and α.
33. Use the Choleski method to calculate the decomposition A = LLT for
4 −1 0 0 0
− 1 4 − 1 0 0
A= 0 −1 4 −1 0
0 0 − 1 4 − 1
0 0 0 − 1 4
1
(A + u vT)-1 = A-1 - A-1u vT A-1
1+λ
where λ = vT A-1u. State the conditions under which it is valid.
(b) Consider solving (A + u vT)x = b. Use the above result to establish that the solution can be
obtained by solving two auxiliary systems
A y = b and A z = u
v.y
and calculating x using the formula x = y - z.
1+(v.z)
[These formulas are useful in the Broyden’s method or other methods that solve nonlinear
equations by rank one updating of the Jacobian matrix].
(a) Show that the Gauss-Seidel Method for solving Ax = b may be written as
xi(k+1) = xi(k)+(Ri(k)/aii),
i-1 n
Ri(k) = bi - aijxj(k+1) - aijxj(k) ; ( i = 1,....,n)
j=1 j=i
Use this form to solve the system Ax = b and tabulate how the residual changes with each iteration.
Take
4 −1 0 0 0
− 1 4 − 1 0 0 10
A = 0 −1 4 −1 0 ;
b= 0 .
0 0 − 1 4 − 1
0
0 0 0 − 1 4 0
and use the initial guess xo = 0.
where ω is the relaxation parameter ( 1< ω < 2), which when selected properly can accelerate the
converge of the iteration. Solve the above system by the SOR method with various values of ω
(1.00, 1.02, 1.05, 1.10, 1.20 and 1.50) and compare the number of iterations needed for obtaining
the solution to four decimal accuracy (absolute error).