Help Topics
Concepts
Package
Class
The Genetic Search Agent panel provides these options:
- Class Name
- Enter the name of a class that represents the population members
and implements the genetic operators and fitness values. Specify it
in package form, such as
com.ibm.package.subpackage.classname
The class should
inherit from the AbleGeneticObject. Pressing the Set button
will cause the class to be introspected and populate the Genetic
Operators and Fitness Values list. The class must be in your
CLASSPATH.
- Population Size
- Enter the size of the population for each generation.
- No duplicates
- Select this checkbox to insure that
only one population member will have a
given chromosome configuration.
- Population Size
- Enter the the number of population members to replace in each generation.
- Normalized fitness
- Select this checkbox to use linear-normalized fitness values
to select individuals for reproduction. If not selected,
raw fitness values are used.
- Maximum Passes
- Enter the maximum number of generations to create during
evolution.
Evolution will stop when either the Maximum Passes or
Fitness Threshold value is reached.
- Fitness Threshold
- When any population member reaches the fitness value,
evolution has progressed sufficiently and will stop.
Evolution will stop when either the Maximum Passes or
Fitness Threshold value is reached.
- Genetic Operators and Fitness Values
- This lists the operators available for use from
this GeneticObject class as well as the probabilities
for their selection.
- Crossover Rate
- Enter the rate that a crossover operator will actually
be applied once the operator is selected.
This is a decimal number between 0 and 1.
- Mutation Rate
- Enter the rate that a bit or gene will be mutated once
the mutate operator is selected.
This is a decimal number between 0 and 1.
The Genetic Search Agent panel is used to create generations until
one member satisfies the desired fitness level, or the number of
generations created reaches a predefined limit.
Steps in using the panel include:
- Enter the name of the class that extends the AbleGeneticObject.
Sample objects are provided in the
com.ibm.able.examples.genetic
package. Providing a class name enables the Set button.
- Press the Set button to read the Genetic Operators and
Fitness Values from the class. If the class is read
successfully, the Initialize button should be enabled.
- Select any other options to use before initializing the agent by
pressing the Initialize button. This will create
the necessary beans and connections between the beans.
- Edit the Genetic Operators as needed to specify the probability
that they will be selected during evolution.
- Set the Crossover and Mutation Rates as desired.
- Press the Start Evolving button to begin creating
generations and evaluating their members for fitness.
Once evolution has begun, this button will toggle to
Stop Evolving. Evolution will continue until the toggled button
is pressed, the number of generations created reaches the
Maximum Passes value, or a population member reaches the
specified Fitness Threshold.
- You may wish to press the Stop Evolving button,
change a property or two, and start again. Press the
Reset Beans button to re-intialize the
population before starting again if you wish.