Package adams.opt.genetic

Class Hermione

  • All Implemented Interfaces:
    Destroyable, GlobalInfoSupporter, LoggingLevelHandler, LoggingSupporter, OptionHandler, Pausable, Randomizable, ShallowCopySupporter<AbstractGeneticAlgorithm>, SizeOfHandler, Stoppable, StoppableWithFeedback, ThreadLimiter, GeneticFitnessChangeNotifier, Serializable
    public class Hermione
extends AbstractClassifierBasedGeneticAlgorithmWithSecondEvaluation
    Hermione. 

    -logging-level <OFF|SEVERE|WARNING|INFO|CONFIG|FINE|FINER|FINEST> (property: loggingLevel)
    The logging level for outputting errors and debugging output.
    default: WARNING
 
    -num-threads <int> (property: numThreads)
    The number of threads to use for executing the jobs; use -1 for all available 
    cores.
    default: -1
    minimum: -1
 
    -num-chrom <int> (property: numChrom)
    The number of chromosomes, ie, the population size.
    default: 50
 
    -stopping-criterion <adams.genetic.stopping.AbstractStoppingCriterion> (property: stoppingCriterion)
    The stopping criterion to use.
    default: adams.genetic.stopping.MaxIterations
 
    -seed <long> (property: seed)
    The seed value for the random number generator.
    default: 1
 
    -initial-setups-provider <adams.genetic.initialsetups.AbstractInitialSetupsProvider> (property: initialSetupsProvider)
    The initial setups provider to use.
    default: adams.genetic.initialsetups.EmptyInitialSetupsProvider
 
    -initial-weights <java.lang.String> (property: initialWeights)
    The initial weights to use, rather than random ones (string of 0s and 1s
    ).
    default: 
 
    -favor-zeroes <boolean> (property: favorZeroes)
    Whether to favor 0s instead of 1s.
    default: false
 
    -best <java.lang.String> (property: bestRange)
    The range of the best attributes.
    default: -none-
 
    -notify <int> (property: notificationInterval)
    The time interval in seconds after which notification events about changes 
    in the fitness can be sent (-1 = never send notifications; 0 = whenever 
    a change occurs).
    default: -1
 
    -bits-per-gene <int> (property: bitsPerGene)
    The number of bits per gene to use.
    default: 1
    minimum: 1
 
    -class <adams.data.weka.WekaAttributeIndex> (property: classIndex)
    The class index of the dataset, in case no class attribute is set.
    default: last
    example: An index is a number starting with 1; apart from attribute names (case-sensitive), the following placeholders can be used as well: first, second, third, last_2, last_1, last; numeric indices can be enforced by preceding them with '#' (eg '#12'); attribute names can be surrounded by double quotes.
 
    -folds <int> (property: folds)
    The number of folds to use in cross-validation.
    default: 10
    minimum: 2
 
    -cv-seed <int> (property: crossValidationSeed)
    The seed value for cross-validation.
    default: 55
 
    -classifier <weka.classifiers.Classifier> (property: classifier)
    The classifier to use.
    default: weka.classifiers.rules.ZeroR
 
    -measure <CC|RMSE|RRSE|MAE|RAE|ACC> (property: measure)
    The measure used for evaluating the fitness.
    default: RMSE
 
    -output-dir <adams.core.io.PlaceholderDirectory> (property: outputDirectory)
    The directory for storing the generated ARFF files.
    default: ${CWD}
 
    -output-type <NONE|SETUP|DATA|ALL> (property: outputType)
    The type of output to generate.
    default: SETUP
 
    -output-prefix-type <NONE|RELATION|SUPPLIED> (property: outputPrefixType)
    The type of prefix to use for the output.
    default: NONE
 
    -supplied-prefix <java.lang.String> (property: suppliedPrefix)
    The prefix to use in case of SUPPLIED.
    default: 
 
    -use-second-evaluation <boolean> (property: useSecondEvaluation)
    If enabled, a second evaluation is performed using the separate folds and 
    seed.
    default: false
 
    -second-folds <int> (property: secondFolds)
    The number of folds to use in cross-validation (second evaluation).
    default: 10
    minimum: 2
 
    -second-cv-seed <int> (property: secondCrossValidationSeed)
    The seed value for cross-validation (second evaluation).
    default: 42
 
    -handler <adams.core.discovery.AbstractGeneticDiscoveryHandler> [-handler ...] (property: handlers)
    The discovery handlers to use.
    default:
    Author:
    Dale (dale at cs dot waikato dot ac dot nz), FracPete (fracpete at waikato dot ac dot nz)
    See Also:
    Serialized Form

    • Constructor Detail

      • Hermione

        public Hermione()

    • Method Detail

      • setHandlers

        public void setHandlers​(AbstractGeneticDiscoveryHandler[] value)
        Sets the discovery handlers to use.
        Parameters:
        value - the classifier

      • handlersTipText

        public String handlersTipText()
        Returns the tip text for this property.
        Returns:
        tip text for this property suitable for displaying in the GUI or for listing the options.

      • newJob

        protected Hermione.HermioneJob newJob​(int chromosome,
                                      int[] w,
                                      weka.core.Instances data,
                                      weka.core.Instances testData)
        Creates a new Job instance.
        Specified by:
        newJob in class AbstractClassifierBasedGeneticAlgorithm
        Parameters:
        chromosome - the number of chromosomes
        w - the initial weights
        data - the data to use
        testData - the test data to use, null for cross-validation
        Returns:
        the instance

      • stringToIntArray

        public int[] stringToIntArray​(String s)
        Converts the bit string into an int array.
        Parameters:
        s - the string to convert
        Returns:
        the bit array

      • intArrayToString

        public String intArrayToString​(int[] ia)
        Int array of bits to string.
        Parameters:
        ia - the int array
        Returns:
        the generated bit string

      • setBitsForPosition

        public void setBitsForPosition​(int chromosome,
                               int[] weights,
                               List<Integer> starts,
                               List<Integer> numbits,
                               int pos,
                               int[] newWeights)
        Updates the bits.
        Parameters:
        chromosome - the chromosome this is for
        weights - the overall weights
        starts - the starting points
        numbits - the number of bits
        pos - the position
        newWeights - the new weights to set

      • getBitsForPosition

        public int[] getBitsForPosition​(int[] weights,
                                List<Integer> starts,
                                List<Integer> numbits,
                                int pos)
        get bit array for parameter at pos.
        Parameters:
        weights - the overall weights
        starts - the starting points
        numbits - the number of bits
        pos - the position
        Returns:
        the weights subset

      • generateClassifier

        public weka.classifiers.Classifier generateClassifier​(int chromosome,
                                                      int[] weights)
        Generate the classifier from current bit array
        Parameters:
        chromosome - the chromosome index
        weights - the bit array
        Returns:
        the classifier