edu.udo.cs.yale.operator.features.construction
Class DirectedGGA

java.lang.Object
  edu.udo.cs.yale.operator.Operator
      edu.udo.cs.yale.operator.OperatorChain
          edu.udo.cs.yale.operator.features.FeatureOperator
              edu.udo.cs.yale.operator.features.selection.AbstractGeneticAlgorithm
                  edu.udo.cs.yale.operator.features.construction.AbstractGeneratingGeneticAlgorithm
                      edu.udo.cs.yale.operator.features.construction.YAGGA
                          edu.udo.cs.yale.operator.features.construction.YAGGA2
                              edu.udo.cs.yale.operator.features.construction.DirectedGGA

All Implemented Interfaces:

ConfigurationListener

public class DirectedGGA
extends YAGGA2

DirectedGGA is an acronym for a Generating Genetic Algorithm which uses probability directed search heuristics to select attributes for generation or removing. Its approach to generating new attributes differs from the original one and is the same as the one of YAGGA.
The (generating) mutation can do one of the following things with different probabilities:

• Probability p/4: Add a newly generated attribute to the feature vector
• Probability p/4: Add a randomly chosen original attribute to the feature vector
• Probability p/2: Remove a randomly chosen attribute from the feature vector

Thus it is guaranteed that the length of the feature vector can both grow and shrink. On average it will keep its original length, unless longer or shorter individuals prove to have a better fitness.
In addition to these mutation heuristics probablilities based on the weights of the attributes are calculated. It is more likely for attributes with a great weight to be selected for generating new attributes. On the other hand the probability for removing an attribute from the example set will decrease for attributes with great weights. This decreases the amount of needed generations drastically.
Another enhancement in comparison to the original GGA is the addition of several generators like the ones for trigonometric or exponential functions. In this way a sinple linear working learning scheme which can deliver weights can be used as inner operator. If this learner can also estimate its performance it is not longer necessary to use a inner cross-validation which also decreases learning time. Such a learner is for example the JMySVMLearner which delivers the xi-alpha performance estimation at least for classification tasks.
. Summarized the advantages of this feature construction algorithm are smaller runtimes and smaller attribute sets as result. These attribute sets increase performance and can be used to explain the models of more complex learning schemes like SVMs. The additional generators allow the construction of features which are not possible by the known kernel functions.
Since this operator does not contain algorithms to extract features from value series, it is restricted to example sets with only single attributes. For (automatic) feature extraction from values series the value series plugin for Yale written by Ingo Mierswa should be used. It is available at http://yale.cs.uni-dortmund.de.

Version:

$Id: DirectedGGA.java,v 1.4 2006/08/03 14:39:36 ingomierswa Exp $

Author:

Ingo Mierswa

Field Summary

Fields inherited from class edu.udo.cs.yale.operator.features.selection.AbstractGeneticAlgorithm

BOLTZMANN_SELECTION, CUT_SELECTION, NON_DOMINATED_SORTING_SELECTION, RANK_SELECTION, ROULETTE_WHEEL, SELECTION_SCHEMES, STOCHASTIC_UNIVERSAL, TOURNAMENT_SELECTION, UNIFORM_SELECTION

Constructor Summary

DirectedGGA(OperatorDescription description)

Method Summary

protected PopulationOperator	getMutationPopulationOperator(ExampleSet eSet) Returns the DirectedGeneratingMutation.
java.util.List<ParameterType>	getParameterTypes() Returns a list of ParameterTypes describing the parameters of this operator.

Methods inherited from class edu.udo.cs.yale.operator.features.construction.YAGGA2

apply, getGenerators, getPreProcessingPopulationOperators

Methods inherited from class edu.udo.cs.yale.operator.features.construction.YAGGA

createInitialPopulation, getGeneratingPopulationOperator

Methods inherited from class edu.udo.cs.yale.operator.features.construction.AbstractGeneratingGeneticAlgorithm

getCrossoverPopulationOperator

Methods inherited from class edu.udo.cs.yale.operator.features.selection.AbstractGeneticAlgorithm

getPostEvaluationPopulationOperators, getPostProcessingPopulationOperators, getPreEvaluationPopulationOperators, solutionGoodEnough

Methods inherited from class edu.udo.cs.yale.operator.features.FeatureOperator

evaluate, evaluate, getCheckForMaximum, getInnerOperatorCondition, getInputClasses, getMaxNumberOfInnerOperators, getMinNumberOfInnerOperators, getNumberOfSteps, getOutputClasses, getRandom, setCheckForMaximum

Methods inherited from class edu.udo.cs.yale.operator.OperatorChain

addAddListener, addOperator, addOperator, checkDeprecations, checkIO, checkNumberOfInnerOperators, checkProperties, clearErrorList, clearStepCounter, cloneOperator, countStep, createExperimentTree, delete, experimentFinished, experimentStarts, getAllInnerOperators, getCurrentStep, getIndexOfOperator, getInnerOperatorForName, getInnerOperatorsXML, getNumberOfAllOperators, getNumberOfChildrensSteps, getNumberOfOperators, getOperator, getOperatorFromAll, getOperators, isEnabled, performAdditionalChecks, removeAddListener, removeOperator, setEnabled, setExperiment, shouldReturnInnerOutput

Methods inherited from class edu.udo.cs.yale.operator.Operator

addError, addValue, addWarning, apply, createExperimentTree, createFromXML, createMarkedExperimentTree, getAddOnlyAdditionalOutput, getApplyCount, getDeliveredOutputClasses, getDeprecationInfo, getDesiredInputClasses, getErrorList, getExperiment, getInput, getInput, getInput, getInputDescription, getIOContainerForInApplyLoopBreakpoint, getName, getOperatorClassName, getOperatorDescription, getParameter, getParameterAsBoolean, getParameterAsColor, getParameterAsDouble, getParameterAsFile, getParameterAsInt, getParameterAsString, getParameterList, getParameters, getParameterType, getParent, getStartTime, getStatus, getUserDescription, getValue, getValues, getXML, hasBreakpoint, hasBreakpoint, hasInput, inApplyLoop, isParameterSet, logMessage, register, remove, rename, resume, setBreakpoint, setInput, setListParameter, setOperatorParameters, setParameter, setParameters, setParent, setUserDescription, toString, writeXML

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

DirectedGGA

public DirectedGGA(OperatorDescription description)

Method Detail

getMutationPopulationOperator

protected PopulationOperator getMutationPopulationOperator(ExampleSet eSet)
                                                    throws OperatorException

Returns the DirectedGeneratingMutation.

Overrides:

getMutationPopulationOperator in class YAGGA2

Throws:

OperatorException

getParameterTypes

public java.util.List<ParameterType> getParameterTypes()

Description copied from class: Operator

Returns a list of ParameterTypes describing the parameters of this operator. The default implementation returns an empty list if no input objects can be retained and special parameters for those input objects which can be prevented from being consumed.

Overrides:

getParameterTypes in class YAGGA2