Scribd
Upload
6 views12 pages

CE601 Num Meth Lecture 7

The document discusses the drawbacks of elimination methods in numerical computations, particularly focusing on round-off errors and ill-conditioned systems. It explains how finite precision in computers leads to inaccuracies and how small changes in coefficients can result in large variations in solutions for ill-conditioned systems. Additionally, it introduces the concept of condition numbers as a measure of sensitivity to changes in system elements, indicating well-conditioned versus ill-conditioned systems.

Uploaded by

S.P. Yadav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views12 pages

CE601 Num Meth Lecture 7

The document discusses the drawbacks of elimination methods in numerical computations, particularly focusing on round-off errors and ill-conditioned systems. It explains how finite precision in computers leads to inaccuracies and how small changes in coefficients can result in large variations in solutions for ill-conditioned systems. Additionally, it introduces the concept of condition numbers as a measure of sensitivity to changes in system elements, indicating well-conditioned versus ill-conditioned systems.

Uploaded by

S.P. Yadav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

CE 601: Numerical Methods

Lecture 7

Course Coordinator:
Dr. Suresh A. Kartha,
Associate Professor,
Department of Civil Engineering,
IIT Guwahati.
Drawback in Elimination Methods
There are various drawbacks while using elimination
methods using computers
o Theoretically, the elimination methods should give
actual solution.
o However in computers we declare or assign
variables with only certain precision (say single
precision, double precision, etc.)
Due to this there are drawbacks like:
o Accumulation of round-off errors.
o Failure in solving ill-conditioned systems.
• Round-off Errors
o Round-off errors are generated by
approximating precision numbers by finite
precision numbers.
o Usually in computer you have single precision
numbers (7 significant digits) and double
precision numbers (14 significant digits).
o The case of effect of round-off errors in
elimination method is demonstrated in next
slide.
0.0003 3 x1 1.0002
1 1 x2 1
Using Gauss elimination method,
0.0003 3 1.0002
( R2 R2 (1 / 0.0003) R1 )
1 1 1
0.0003 3 1.0002
0 9999 3333
x2 0.3333333........
x1 0.666666.........
o But while using computer, you need to use finite precision
numbers.
Accuracy of Precision X2 x1
significant Digits
4 4 decimals 0.3333 1.000
5 5 decimals 0.33333 0.70000
6 6 decimals 0.333333 0.670000
7 7 decimals 0.3333333 0.6670000
o You can see from the above table, how based
on the significant digits and precision the
round-off errors creep.
o You can reduce round-off errors by doing
partial pivoting.
• Ill-conditioned System
Consider the system
x1 x2 2
x1 1.0001x2 2.0001
1 1 2
or,
1 1.0001 2.0001
It's solution is x2 1.0 and x1 1.0.
o if we give a slight perturbation on the
coefficient of x2 in the above equation,
i.e., x1 x2 2
x1 0.9999 x2 2.0001
1 1 2
or,
1 0.9999 2.0001
Now the solutions are x2 1 and x1 3.

o Huge difference in solution for a very slight


changes in the system.
o This is because the system is ill-conditioned.
• An ill-conditioned system is one where small
changes in the numerical values of coefficient
matric [A] or right side vector{b} cause large
changes in the solution vector {x}.
• A well-conditioned system is one which produce
only small change in the solution vector for small
changes in [A] or {b}.
Example of an ill-conditioned system has already
been discussed in our previous class.
• Ill-conditioning of a system is determined by its
condition number.
• Norms
o Measure of magnitude of a matrix, vector,
etc. For [A], {b}, {x} etc, their norms are given
as :
A , b , x etc.
The norm of a matrix or vector shall be
always greater than zero.
i.e., A 0, x 0
If A 0, then [ A] 0.
If k is a scalar quantity, then k[A]=[kA], but
kA k A ,(norm of a scalar is its absolute value)
 For [ A] and [ B],
A B A B
AB A B
 Norm of a vector {x} x
You have, x 1 x i (sum of magnitudes)
1
2 2
x 2
x e
x
i (Euclidean Norm)
x max xi (Maximum magnitude Norm)
1 i n
 For an n n matrix [ A],
n
A1 max aij (Maximum of column sums)
1 j n
i 1
n
A max aij (Maximum of row sums)
1 i n
j 1
1
n n 2
Ae aij2 (Euclidean norm)
i 1 j 1

• Condition Number
o Measure of sensitivity of the system to small
changes in any of its elements.
Recall, [ A]{x} {b}
Now it is clear, b A x (from properties of norm)
Small change { b} in {b} will cause change in {x},say by { x}.
So, new system will be, [ A] {x} { x} {b} { b}
[ A]{ x} { b} ( [ A]{x} {b})
{ x} [ A] 1{ b}
1
x A b
1
b x A x A b ( b A x)
x 1 b
A A
x b
1
where c( A) A A is called condition number.
x b
i.e., c( A)
x b
Smaller values of c(A) -> well conditioning
Larger values of c(A) -> ill conditioning
• Example
1 1
Find condition number of [ A]
1 1.0001
Ae 2.00005

1 1.0001 1 10001 10000


[ A] 10000
1 1.0001 10000 10001
1
A 20000.50002
e
1
c( A) A A 40002
c( A) Very large value Ill-conditioned system.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy