ec.rule.README Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of ecj Show documentation

ECJ, A Java-based Evolutionary Computation Research System. ECJ is a research EC system written in Java. It was designed to be highly flexible, with nearly all classes (and all of their settings) dynamically determined at runtime by a user-provided parameter file. All structures in the system are arranged to be easily modifiable. Even so, the system was designed with an eye toward efficiency. ECJ is developed at George Mason University's ECLab Evolutionary Computation Laboratory. The software has nothing to do with its initials' namesake, Evolutionary Computation Journal. ECJ's sister project is MASON, a multi-agent simulation system which dovetails with ECJ nicely.

The newest version!

This package contains classes for representations of the form of arbitrary
collections of objects.  One common use of such collections would be as
bags of "rules", hence the name.  However there's no reason you couldn't use
any kind of object here.  Also don't be put off by the name "ruleset" -- it's
not a set of rules, it's a collection (a bag, a multiset) of objects, that is,
there's no reason the object can't appear more than one time.

The package is very straightforward.  A RuleIndividual contains a fixed-length
array of RuleSets.  Each RuleSet is an arbitrary-length array of Rules.
A Rule is the superclass of your basic object.

RuleIndividuals must belong to RuleSpecies (or a subclass), which does trivial
checks.  Various RuleSets are associated with a RuleSetConstraints object
in which they can store common information.  Presently RuleSetConstraints
specify:

	- the minimum number of rules allowed in the RuleSet upon initialization
	- the maximum number of rules allowed in the RuleSet upon initialization
or
	- a distribution from which the number of rules is drawn

Additionally:

	- The probability that new rules will be added upon mutation
	- The probability that new rules will be deleted upon mutation
	- The probability the rules will be randomized in order upon mutation

You're welcome to modify this as you see fit of course.  Rules are likewise
associated with RuleConstraints in which they can store common information
(presently nothing is stored).

To set up the constraints objects, the rule package requires that your
initializer be a subclass of RuleInitializer.

The 'breed' directory contains two breeding pipelines which might be of use.
RuleCrossover trades rules among rulesets using the following algorithm:

	For i from 1 to the number of rulesets
		r1 = ruleset i in individual A
		r2 = ruleset i in individual B
		For each rule in r1
			With a given probability,
				mark that rule to go to r2
		For each rule in r2
			With a given probability,
				mark that rule to go to r1
		Exchange marked rules between r1 and r2

You specify the rule exchange probability.  The RuleMutation operator simply
calls mutateRules() on all the rulesets.  The default mutateRules() function
first calls mutate() on al the rules in the ruleset.  It then deletes rules from
the ruleset with a certain probability.  It then adds new rules to the ruleset
with a certain probability.  Finally it randomizes the rule order of the ruleset
with a certain probability.