Assignment #1 MACM 316 Gaussian Elimination & Prerequisite Problems

please respect page limits & practice the Guidelines for Written Assignments.
submit problem A) to the homework box # 15 by 11:50pm, Wednesday 12 September.
your scripts/codes must be uploaded to the Canvas assignment by this same time.
problems B) & C) are not to be submitted.
the related quiz will be given in lecture, Monday 15 September.
remember that Canvas discussion is an open forum.
acknowledge collaborations & assistance from colleagues/TAs/instructor.
A) Computing Problem: Gaussian Elimination (1 page report) The Matlab demo in the
rst lecture showed that nite-precision leads to tiny errors for Gaussian elimination (GE)
with small matrix sizes. A larger matrix seemed to result in larger errors. To quantify the
error of a GE calculation, dene an error vector, , as the dierence between the computed
and the exact solution vectors
x
0
=
_
_
_
_
_
_
_
_
1
.
.
.
1
.
.
.
1
_
_
_
_
_
_
_
_
; x
1
=
_
_
_
_
_
_
_
_
1 +
1
.
.
.
1 +
j
.
.
.
1 +
N
_
_
_
_
_
_
_
_
; = x
1
x
0
=
_
_
_
_
_
_
_
_

1
.
.
.

j
.
.
.

N
_
_
_
_
_
_
_
_
.
A scalar error for a single GE calculation can be dened by the worst of the component
errors, that is, the maximum of the absolute values |
j
|
= max
j
|
j
| .
Since this error value depends on the matrix used, a typical error value might be obtained
by averaging many such values
(k)
obtained from suciently many k = 1 . . . M
N
dierent
random matrices
E
N
=
1
M
N
M
N

1

(k)
.
Dene this E
N
as your computed value of the typical error for GE with matrices with size N.
Your investigation here is to estimate the size of matrix, N

, at which your typical

error for GE, E
N
1, becomes the same size as the component of the exact
solution x
0
!
Of course, your laptop or the lab PC will be nowhere large enough to test this result. But you
can extrapolate an estimate by obtaining typical errors E
N
for several dierent values of N.
Make a plot of points (log
10
N, log
10
E
N
). The use of base 10 logarithms makes it convenient
to interpret this plot in terms of decimal places. You will have to decide how many dierent
values of N and how many samples M
N
are needed to generate points from which a reasonable
extrapolation can be obtained.
Explain your conclusions in a one-page report that includes your error plot. Be sure to follow
the Guidelines for Written Assignments on the Canvas webpage.
B1) Linear Algebra: Gaussian Elimination
Gaussian elimination, or row reduction, is a systematic way to solve a set of linear equations
(Lay, s1.1 & 1.2). Solve the following 3 3 linear system
x 2y + z = 0
2y 8z = 8
4x + 5y + 9z = 9
.
B2) Linear Algebra: Eigenvalues & Eigenvectors
Find the eigenvalues {
1
,
2
,
3
} and corresponding eigenvectors { w
1
, w
2
, w
3
} for the matrix
A =
_
_
4 0 1
2 1 0
2 0 1
_
_
.
Give the linear system that determines the linear combination
a
1
w
1
+ a
2
w
2
+ a
3
w
3
=
_
_
3
0
3
_
_
,
without solving for the a
j
coecients, how can you determine whether this system is solvable?
B3) Linear Algebra: Gram-Schmidt Procedure
Look up the Gram-Schmidt procedure for producing an orthogonal basis. Apply it to the
vector subspace spanned by the vectors
_

_
_
_
_
_
1
1
1
1
_
_
_
_
,
_
_
_
_
0
1
1
1
_
_
_
_
,
_
_
_
_
0
0
1
1
_
_
_
_
_

_
.
C1) Calculus: 1D Taylor Series
Derive the rst three terms of the Taylor series for
f(x) = ln x = c
0
+ c
1
(x 2) + c
2
(x 2)
2
+ R(x)
about the value x = 2, where R(x) is the remainder term (Stewart, s11.10). Give the Taylor
inequality for the remainder size |R(x)| on the interval 1 x 3.
C2) Calculus: 2D Linear Approximation
Derive the best linear approximation (Stewart, s14.4) of the function z(x, y) = xe
xy
valid near
the point (x, y, z) = (2, 0, 2).