Course Assignments for CSCE/Math 441

Thomas Shores
Department of Mathematics
University of Nebraska
Fall 2009
Last update: 08/25/09
Note: Unless otherwise stated, it is always permissible to use the course software (Matlab,
Octave) for calculations. Also, Maple is acceptible for some tasks such as checking (or doing)
your algebraic calculations. As a general rule, you are expected to show your work. In
particular, if you use software to solve a problem, you are expected to provide a transcript
or some output and paste it into a document. If an exercise is a programming problem,
provide some sample output to show the correctness of your program. Typed documents
are preferred; hand written copy will be accepted if it is neatly written. Unless otherwise
stated, hardcopy is the rule. It is expected that co-collaborators and other sources for the
homework will be duly acknowledged. Problems that are to be worked by individuals without
collaboration will be marked (I). Each assignment will be worth about 40 points.

Assignment 1
Points: 35
Due: Friday, September 11
Do Exercises 1-8 in the le VectorSpaces441.pdf which can be found in the Course Mate-
rials folder.
Assignment Closed

Assignment 2
Points: 35
Due: Friday, October 2
Do Exercises 1.5, 2.7, 3.9 and 4.8 from textbook. Do these non-text exercises:
N1: With the assumptions and notation of Theorem 3.1 in the text, if one has a good
∗
estimate of of d (f ), one can nd a lower bound for kXk. Show this lower bound. Then
assume that Chebyshev interpolation gives a good estimate to the best approximation to
f ∈ C [−5, 5] from B = P20 and estimate d∗ (f ) for f = 1/(1 + x2 ) and one other function of
your choice. Use these to give lower bounds for kXk, where X is the interpolation operator
using 12 equally spaced interpolation points.
N2: Write a Matlab function with calling form Lnorm(a,b,x), where inputs a and b are
left and right endpoints of an interval containing the coordinates of the input vector x, and
the output is an approximation to the norm of the Lagrange interpolation operator with
interpolation points in the vector x.
Assignment Closed

Assignment 3
Points: 40
Due: Friday, October 16
Do Exercises 5.5, 5.8, 6.2 from textbook. Do these non-text exercises:
1
 2

N3: Write a Matlab function with the calling form HCpp(x,f,fp), where this function
returns a PP structure representing the Hermite cubic p.p. with node row vector x, corre-
sponding function values in the row vector f and derivative values in the row vector fp. You
should follow the guidelines of the le PPfcns.m that is found in Course Materials. Validate
with some calculations with the Runge's example and several choices of knots and include
printout for the case of three knots.
N4: It is claimed in the notes that the error of Hermite p.p. interpolation is O (h4 ), where
h is the largest distance between knots, so that, if the error estimate is reasonably sharp,
halving h should reduce the error by a factor of about 16. Verify this with Runge's example
on the interval [0, 5] and equally spaced knots with h = 1, h = 1/2 and h = 1/4.
Assignment Closed
Assignment 4
Points: 40
Due: Tuesday, November 17
Do Exercise 12.2 from textbook. (Note that for this exercise you must nd the rst three
orthogonal polynomials with respect to the weight function w (x) = x on the interval [0, 1].
Then nd the zeros of the quadratic for the nodes x0 , x1 and use Formula (12.11) in the text
to compute the weights w0 , w1 .)
N5: Calculate a Pade approximation for f (x) = ex of total degree 3 = N = n + m, with
n=2 and m = 1. (See the ClassroomNotes441.pdf le for an explanation of the notation.)
Next nd the best cubic approximation to f (x) on [−1, 1]. Find the (approximate) max
norm of the error with Matlab. Which is better? (See our ClasroomNotes le for more
information on Pade approximations.)
N6:Determine by Matlab experiment how many function evaluations are required to
R1 x
approximate
−1
e sin (πx) dx to an accuracy as good as one obtains with a 7 point Gaussian
quadrature method, if one uses a trapezoidal method with equally spaced nodes. (Exact
answer: ((e − 1/e) / (π + 1/π).)
Assignment Closed
Assignment 5
Points: 40
Due: Wednesday, December 9
N7: Verify by direct hand calculation that FN−1 = N F N in the case that N = 3.
N8: Make out a table with two columns. On the left write the binary format of integers
0 through 3, then in the second column reect the binary form and convert the result to
decimal numbers. Repeat this with integers 0 through 7. What does this tell you about the
ordering for the indices of the FFT?
N9: Compare the op count for computing the product of a polynomial of degree 1000
and a polynomial of degree 2 by direct calculation and by the FFT method. Who wins?
N9: Write a Matlab routine that implements FFTsort.
N10: Use your FFTsort program and FFTransform to conrm in Matlab the claim that

(yk ) = N FN (Y n ) with a few calculations involving data vectors of length 4, 8, 16.