Practical Implementation Issues
The next set of slides will deal with some issues
encountered in the practical implementation of a
genetic algorithm.
Specifically constraint handling and convergence
criteria.
docsity.com
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Practical Implementation Issues-Stochastic Process-Lecture Slides, Slides of Stochastic Processes
Main topics for this course are Stochastic process, random variables, linear congruent generators, pdfs and cdfs, rejection method, metropolis methods, sampling techniques, random walks and genetic algorithm. This lecture includes: Practical, Implementation, Issues, Genetic, Algorithm, Constraint, Handling, COnvergence, Design, Penalty, Term, Quality, Encode
Typology: Slides
2011/2012
1 / 20
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Practical Implementation Issues-Stochastic Process-Lecture Slides and more Slides Stochastic Processes in PDF only on Docsity!
Practical Implementation Issues The next set of slides will deal with some issuesencountered in the practical implementation of agenetic algorithm. Specifically constraint handling and convergencecriteria.
Constraint Handling What if we have a constrained problem? Typically would use a penalty function to worsenthe fitness of any designs which violateconstraints.
Convergence Common ways of determining convergence: 1.^
No change in best quality design over chosennumber of gens.
2.^
No change in average quality of population ofchosen number of gens.
3.^
Set a maximum allowable number ofgenerations.
4.^
Set a maximum allowable number of objectivefunction evaluations.
Putting It All Together
begin
t = 0 initialize P(t) evaluate P(t) while
(not converged)
do
t = t + 1 select P(t) from P(t-1) alter P(t) (variation operators) evaluate P(t) end do while end
Example According to our 5 basic components, what shouldour first step be? Considering the nature of our problem, is there anencoding scheme that we think may work well? How can we encode solutions to our problem inthis way?
Example We have decided on a vector of bit encodingbecause each of our switches has 2 states (on andoff).
So each of our encoded strings will look like this:
[ {0,1}, {0,1}, {0,1}, {0,1}, {0,1} ]
An example of which would be:
[ 0, 1, 0, 1, 0 ]
Example Now what? We must decide how we are going to evaluate our designfitness.
This is not (necessarily) the objective function value. In our case, we will consider the output (payoff) of ourblack box to be the design fitness (We will treat the switchsettings as a binary number and our fitness will be thesquare of the decimal equivalent).
Example At this point, we have to decide how to implementour selection and variation operators. Selection As mentioned when presenting selectionstrategies, we will use the f/f
avg^
approach.
Example Variation – Mutation We will use random bit mutation which in this caseis exactly the same as random design variablereassignment since each of our design variables isin itself a bit.
Example Parameters We need to decide on values for the parameters ofour algorithm.
ie: mutation and crossover rate,
population size, convergence criteria (note that weare not going to run this optimization tocompletion so we will not explicitly worry aboutconvergence).
Example Initial Population – Gen. #0 (recall f = x
string
X^
f^
f/favg
cop.
avg
Example Generation #1 - variation^ string
after cross. after mutation
Example Generation #2 - variation^ string
after cross. after mutation
Example Generation #2 - evaluation + selection for 3:^ string
X^
f^
f/favg
cop.
avg