MATH2019 (2019–2020)
Introduction to Scientific Computation
1st Semester — Lecture 5
(Linear Systems: Direct Methods II)
PROBLEM SHEET
Reading material: Burden, Faires & Burden 2016 (10th ed.)
• Section 6.2
In-class problems
Problem 1 (Gaussian elimination: Round-off errors)
Consider the augmented matrix
0.04 50 50.4
à =
8 −40 40
a Perform Gaussian elimination (with backward substitution) to obtain the so-
lution (x1 = 10, x2 = 1).
b Perform Gaussian elimination to obtain the solution, but now using
3-digit arithmetic with rounding.
Problem 2 (Gaussian elimination: Pivoting strategies with round-off)
a Which pivoting strategy works well for the 1st system? (Assume 3-digit
arithmetic with rounding.)
0.04 50 50.4
à =
8 −40 40
b Which pivoting strategy works well for the 2nd system? (Assume 3-digit
arithmetic with rounding.)
40 50000 50400
à =
8 −40 40
Additional problems
Problem 3 (Gaussian elimination: Round-off errors)
Previously we considered the augmented matrix
0.04 50 50.4
à =
8 −40 40
School of Mathematical Sciences — Version: 1st November, 2019 — KGvdZ Page 1 of 3
Now, consider the augmented matrix
40 50000 50400
à =
8 −40 40
Is there a difference for the exact solution? Is there a difference when doing
Gaussian elimination to obtain the solution, but using 3-digit arithmetic with
rounding?
Problem 4 (Gaussian elimination: Pivoting strategies, exact arithmetic)
Burden, Faires & Burden 2016 (10th ed.), Exercise Set 6.2, Exercise 1(a,b), 3,
5, 7:
Consider the following linear systems:
a Find the row interchanges that are required to solve each system using
Algorithm 6.1 (i.e., standard pivoting). You can assume exact arithmetic.
b Find the row interchanges that are required to solve each system using
partial pivoting. You can assume exact arithmetic.
c Find the row interchanges that are required to solve each system using
scaled partial pivoting. You can assume exact arithmetic.
d Find the row interchanges that are required to solve each system using
complete pivoting. You can assume exact arithmetic.
Problem 5 (Finite-digit arithmetic: Examples)
Burden, Faires & Burden 2016 (10th ed.), Exercise Set 1.2, Exercise 5(a-c):
Perform the following computations (i) exactly, (ii) using three-digit chopping
arithmetic, and (iii) using three-digit rounding arithmetic. (iv) Compute relative
y−f l(y)
errors y for the exact answers y and their approximations f l(y) at (ii)
and (iii).
4 1
a. +
5 3
4 1
b. ·
5 3
1 3 3
c. − +
3 11 20
Problem 6 (Gaussian elimination: Finite-digit arithmetic)
Burden, Faires & Burden 2016 (10th ed.), Exercise Set 6.1, Exercise 3(a):
Use Gaussian elimination with backward substitution and two-digit rounding
MATH2019 PROBLEM SHEET — 1st Semester — Lecture 5 (Linear Systems: Direct Methods II) Page 2 of 3
arithmetic to solve the following linear system (do not reorder the equations):
4x1 − x2 + x3 = 8 ,
2x1 + 5x2 + 2x3 = 3 ,
x1 + 2x2 + 4x3 = 11 .
(The exact solution is x1 = 1, x2 = −1, x3 = 3.)
Problem 7 (Gaussian elimination: Pivoting strategies with round-off)
Burden, Faires & Burden 2016 (10th ed.), Exercise Set 6.2, Exercise 11, 15, 9,
13:
a Use Gaussian elimination and three-digit rounding arithmetic to solve the
following linear system, and compare the approximation to the actual solu-
tion (x1 = 10, x2 = 1):
0.03x1 + 58.9x2 = 59.2 ,
5.31x1 − 6.10x2 = 47.0 .
b Repeat Part a using Gaussian elimination with partial pivoting.
c Repeat Part a using three-digit chopping arithmetic.
d Repeat Part a using Gaussian elimination with partial pivoting, and three-
digit chopping arithmetic.
Problem 8 (Pivoting strategies: Counting number of comparisons)
a To find cp = max{c1 , c2 , . . . , cm }, show that the number of comparisons
between two numbers equals m − 1.
b Partial pivoting computes
|api | = max |aji | for each i = 1, 2, . . . , n − 1.
1≤j≤n
Compute the total number of comparisons carried out using partial pivoting.
c Complete pivoting computes
|apq | = max |ajk | for each i = 1, 2, . . . , n − 1.
1≤j≤n
1≤k≤n
Compute the total number of comparisons carried out using complete piv-
oting.
MATH2019 PROBLEM SHEET — 1st Semester — Lecture 5 (Linear Systems: Direct Methods II) Page 3 of 3
