Wyższa Szkoła Zarządzania i Bankowości
LECTURESNOTEBOOKSPACKAGES
4 Number Representation and Error Propagation
4.3 Conditioning

1 Introduction to Simulation and Modeling
2 Discrete Medeling (L-Systems)
3 Population Dynamics
4 Number Representation and Error Propagation
4.1 Number Representation
4.2 Error Propagation
4.3 Conditioning
5 Modeling with Random Numbers
6 Heat Transfer in a Rod (Connection Mathematica and C: MathLink)
7 Special Topics in Stochastic Finance
8 Appendix: Introduction to Mathematica
9 Population Dynamics in Vensim®PLE
     
 

Conditioning

Even when calculations are performed exactly, there are situations for which it can be difficult or impossible to obtain a correct answer. Chaotic systems are extremely sensitive to small changes in the start conditions. A well-known example of such a system is weather forecasting. The chaotic nature of the system imposes a limit on the period at which the weather can be forecasted with some degree of confidentiality (this period is about 15 days!).
The sensitivity of a system to changes in the start conditions is expressed in terms of the condition number, which is defined as

[Graphics:Images/nb4_gr_286.gif].

The condition number determines whether a system is sensitive to changes in the start conditions. A large condition number makes a system ill-conditioned and sensitive to small changes. A small condition number on the other hand defines a well-conditioned system where the system behaves well to small changes.

Condition number of linear functions

Assume that the approximated result for [Graphics:Images/nb4_gr_287.gif] follows from using the approximated value of [Graphics:Images/nb4_gr_288.gif], for some linear function [Graphics:Images/nb4_gr_289.gif]. (Whenever [Graphics:Images/nb4_gr_290.gif] is not a linear function take the first order Taylor expansion.) It was shown in the previous section that for any number the computed or approximated value can be written as

[Graphics:Images/nb4_gr_291.gif].

Applying this to [Graphics:Images/nb4_gr_292.gif] gives

[Graphics:Images/nb4_gr_293.gif],             (1)

where [Graphics:Images/nb4_gr_294.gif] is the approximated value to [Graphics:Images/nb4_gr_295.gif]. The assumption that [Graphics:Images/nb4_gr_296.gif] follows from the approximated value for [Graphics:Images/nb4_gr_297.gif] can be written as

[Graphics:Images/nb4_gr_298.gif].             (2)

Combining (1) and (2) leads to

[Graphics:Images/nb4_gr_299.gif]  (3)

The fraction [Graphics:Images/nb4_gr_300.gif] is a measure for the sensitivity of this system. If the derivative of [Graphics:Images/nb4_gr_301.gif] exists in [Graphics:Images/nb4_gr_302.gif] the following is true

[Graphics:Images/nb4_gr_303.gif].

Together with (3) this gives

[Graphics:Images/nb4_gr_304.gif],

as [Graphics:Images/nb4_gr_305.gif]. This result gives us a measure for the relative sensitivity of [Graphics:Images/nb4_gr_306.gif] for small variations in [Graphics:Images/nb4_gr_307.gif]. By taking the absolute value the condition number of [Graphics:Images/nb4_gr_308.gif] for [Graphics:Images/nb4_gr_309.gif] is obtained:

[Graphics:Images/nb4_gr_310.gif].

Take for example the function [Graphics:Images/nb4_gr_311.gif]. The relative derivative is given by [Graphics:Images/nb4_gr_312.gif]. The condition number is therefore [Graphics:Images/nb4_gr_313.gif]. This implies that for  [Graphics:Images/nb4_gr_314.gif] the function is sensitive to small changes in its arguments. For [Graphics:Images/nb4_gr_315.gif] on the other hand, small variations in the arguments will have only small effects.

REQUIRED
IV.2a
Give the condition number for an argument [Graphics:Images/nb4_gr_316.gif] for the function [Graphics:Images/nb4_gr_317.gif]. Also discuss when the function is sensitive to small changes in its arguments.


 
     
  Lectures | Notebooks | Packages

 
  Copyright © 2003 Wyższa Szkoła Zarządzania i Bankowości. Wszystkie prawa zastrzeżone
webmaster@wszib.edu.pl