开源可二次开发的遗传算法包jgap

3.3.3 ----------------------------------------------------------- * Simplified the MinimizingMakeChange example * Added new class DummyGridClientMediator * Added new class MessageContext * Added new class BasicContext * Added new class PersistableObject * Added new base class NaturalSelectorExt * Added new class DateKit * Added default constructor for class NaturalSelector * Did a lot of architectural improvements in JGAPClientGP * Added a bunch of attributes to JGAPResultGP * Added a bunch of attributes to JGAPRequestGP * JGAPWorkerGP: measure duration of work * Added methods to IGridClientMediator * Considered age in BestChromosomesSelector * Inherited BestChromosomesSelector from NaturalSelectorExt * Inherited WeightedRouletteSelector from NaturalSelectorExt * Inherited TournamentSelector from NaturalSelectorExt * Inherited ThresholdSelector from NaturalSelectorExt * Enhanced Configuration.clone() * Fixed bug in GPConfiguration.clone() * Fixed bug in class MathProblem, avoiding evolution * Fixed bug in GPPopulation.determineFittestProgram() * GPConfiguration: added method setFitnessEvaluator * Improved GP TournamentSelector * Introduced new class StaticPopulationCreator * GPProgramBase: Replaced -1 with constant * GPPopulation: added methods setGPPrograms and copyGPPrograms * GPFitnessFunction: caught exception in fitness determination that could result from ensureUniqueness * Enhanced conf. settings in CrossoverOperator * Enhanced conf. settings in DefaultConfiguration * Refactored GABreeder to better design * Fixed bug in BreederBase with multiple selectors * JGAPResultGP: added support for generic data * MasterInfo: added preset constructor; added username, GUID * FileKit: added getTemDir() * Enhanced Javadoc, fixed spelling errors * Added some package.html description files * Added a note about Xerces in lib/readme.txt 3.3.2 ----------------------------------------------------------- * Added new class RangedSwappingMutationOperator, thanx to Florian * Added package description to equalDistribution package * Added method setMutationRate to class MutationOperator * Fixed bug in GABreeder with preservation of fittest, thanx to Oscar * Fixed bug in Population with preservation of fittest, thanx to Francisco * Enhanced CrossoverOperator: full x-over selected by default; allow x-over of new age chroms (thanx to Zahir) * Introduced trove maps to CachedFitnessFunction * Enhanced CachedFitnessFunction, thanx to Tobias * Added new unit tests * Added one-file PDF documentation, thanx to Denis * Equipped MapGene and SetGene with fast trove4j maps * Made all tests for MapGene and SetGene pass * Enhanced Javadoc, fixed spelling errors 3.3.1 ----------------------------------------------------------- * Introduced mutation of Genetic Programming commands * Utilized XStream library for serialization issues * Utilized trove4J: fast and consistent map replacement * Integrated Robocode 1.5 with JGAP * Added cloning capabilities to many RobocodeJGAP commands * Fitness value of original chromosome is set to clone when cloning the chromosome * Enabled cloning of application data in genes * Moved method niceMemory from GPGenotype to SystemKit * Made class IComplexCommand serializable * Added execute_object() to ForLoop * Fixed output of generation number in GPGenotype.evolve() * Briefer log messages in GPGenotype * Added default constructor to GPFitnessFunction * Renamed all auxiliary test classes from xxxForTest to xxxForTesting * Fixed Maven pom file (thanx to Fabian) * Enhanced Genotype tests to check for unique chromosomes 3.3 ----------------------------------------------------------- * Integrated Robocode 1.4.8 with JGAP * Added persistent representations of GPPopulation, GPProgram, ProgramChromosome and CommandGene * Enhanced interface IProgramCreator * GPProgram: added a growOrFull method for initializing programs with given chromosomes * CommandGene: added service method ensureUniqueness * Now infinity is not accepted as a fitness value * New method getAssignableFromClass in BaseGPChromosome * New interface IComplexCommand for declaring GP commands as complex * Enhanced DefaultGPFitnessEvaluator: capable of handling IllegalStateException's (for wrong programs) * Enhanced GPProgram: capable of handling IllegalStateException's (for wrong programs) * Enhanced GPGenotype.outputSOlution() * Fixed serious bug with GP mutation of functions * Fixed bug with GP evolution (probability fix) * Fixed Java 6 bug with MathProblem * Fixed bug in GPProblem.createTree * Fixed bug in DynamicMutationExample * Enhanced logging of GP information * Enhanced Javadoc * Added some unit tests 3.2.2 ----------------------------------------------------------- * Enhanced GP evolution: prototype program always recognized in first evolution round * Added IPopulationCreator and DefaultPopulationCreator: allow for presetting the GP population at start-up * Added IProgramCreator and DefaultProgramCreator: allow for sophisticated custom initialization of GP programs (IGPProgram) * Added convenient constructor to ProgramChromosome * Chromosome can be configured to always calculate fitness value (instead of caching only) * Added getGenes() to class CompositeGene * Added some few unit tests * Added missing Javadoc, enhanced Javadoc * Preset keepPopulationSize to true instead of false * Fixed two bugs in MultiObjectiveExample * Fixed small bugs in Configuration with locking null valued properties 3.2.1 ----------------------------------------------------------- * Made Robocode example work with newest Robocode version * Enhanced Robocode example a lot * Added Maven pom file * Added custom-initialization mechanism for GP * Introduced log4j * Fixed bug with allTimeBest and cloning (see bug 1744077) * Made INodeValidator serializable * Enhanced IGPChromosome by method getFunctionSet() * Added Java command version to NOP * Enhanced Javadoc a lot * Fixed bug with GABreeder.evolution (bug 1748528) * Added two utility functions to SystemKit * Fixed some few Javadocs * Improved Chromosome.hashCode() * Added some few unit tests 3.2 ----------------------------------------------------------- * Ported JGAP from Java 1.4 to Java 5 * Enhanced architecture BaseChromosome, Chromosome a lot * Introduced breeder (interface IBreeder, base class BreederBase, class GABreeder) to simplify evolution outside/without Genotype --> grid * Enhanced grid logic a lot * Enhanced GP program generation a lot (a low "max nodes" parameter does not lead to program abortion any more; auto-consider NOP if necessary for creating valid programs...) * Enhanced evolution loop * Improved speed of UTF-8 encoding * Unified character encoding and decoding * Introduced CachedFitnessFunction * Enabled GPConfiguration for grid computing * Better ADF support in Genetic Programming * Added GP function ExchangeMemory and others * Added GP program caching to speedup evolution * Added GP TournamentSelector (by Javier and Enrique) * Added complex Robocode example (thanx to Javier and Enrique) * Removed static attribute from get/setGPConfiguration in class GPGenotype * Added matrixed memory to class Culture * Added contains method to class Culture * Added concrete getName() to functions and terminals * Added example with distributed fitness computation * Introduced grid classes IEvolveStrategy and DefaultEvolveStrategy * Made IGPProgram cloneable --> please consider when upgrading your code * Made WeightedRouletteSelector serializable * Made BestChromosomesSelector cloneable * Changed configuration to avoid using unmodifyable collection for genetic operators (as it is not cloneable) * Added job-oriented approach (package org.jgap.impl.job) for grid computing * Introduced grid classes IWorkerRet