Documentation of 'ec.vector.VectorSpecies' Java class
VectorSpecies
Skip navigation links
ec.vector

Class VectorSpecies

  • All Implemented Interfaces:
    Prototype, Setup, java.io.Serializable, java.lang.Cloneable
    Direct Known Subclasses:
    BitVectorSpecies, FloatVectorSpecies, GeneVectorSpecies, IntegerVectorSpecies

    public class VectorSpecies
extends Species
    VectorSpecies is a species which can create VectorIndividuals. Different VectorSpecies are used for different kinds of VectorIndividuals: a plain VectorSpecies is probably only applicable for BitVectorIndividuals.

    VectorSpecies supports the following recombination methods:

    • One-point crossover.
    • Two-point crossover.
    • Uniform crossover - inaccurately called "any-point".
    • Line recombination - children are random points on a line between the two parents.
    • Intermediate recombination - the value of each component of the vector is between the values of that component of the parent vectors.

    Note that BitVectorIndividuals (which use VectorSpecies) and GeneVectorIndividuals (which use GeneVectorSpecies, a subclass of VectorSpecies) do not support Line or Intermediate Recombination.

    Also note that for LongVectorIndividuals, there are certain values that will never be created by line and intermediate recombination, because the recombination is calculated using doubles and then rounded to the nearest long. For large enough values (but still smaller than the maximum long), the difference between one double and the next is greater than one.

    VectorSpecies has three wasy to determine the initial size of the individual:

    • A fixed size.
    • Geometric distribution.
    • Uniform distribution

    If the algorithm used is the geometric distribution, the VectorSpecies starts at a minimum size and continues flipping a coin with a certain "resize probability", increasing the size each time, until the coin comes up tails (fails). The chunk size must be 1 in this case.

    If the algorithm used is the uniform distribution, the VectorSpecies picks a random size between a provided minimum and maximum size, inclusive. The chunk size must be 1 in this case.

    If the size is fixed, then you can also provide a "chunk size" which constrains the locations in which crossover can be performed (only along chunk boundaries). The genome size must be a multiple of the chunk size in this case.

    VectorSpecies also contains a number of parameters guiding how the individual crosses over and mutates.

    Per-Gene and Per-Segment Specification. VectorSpecies and its subclasses specify a lot of parameters, notably mutation and initialization parameters, in one of three ways. We will use the mutation-probability parameter as an example.

    1. Globally for all genes in the genome. This is done by specifying:

      base.mutation-probability
      base.max-gene

      Note: you must provide these values even if you don't use them, as they're used as defaults by #2 and #3 below.

    2. You may provide parameters for genes in segments (regions) along the genome. The idea is to allow you to specify large chunks of genes all having the same parameter features. To do this you must first specify how many segments there are:

      base.num-segments

      The segments then may be defined by either start or end indices of genes. This is controlled by specifying the value of:

      base.segment-type

      ...which can assume the value of start or end, with start being the default. The indices are defined using Java array style, i.e. the first gene has the index of 0, and the last gene has the index of genome-size - 1.

      Using this method, each segment is specified byj...

      base.segment.j.start
      base.segment.j.mutation-probability if segment-type value was chosen as start or by:

      base.segment.j.end
      base.segment.j.mutation-probability if segment-type value is equal to end.

    3. You may parameters for each separate gene. This is done by specifying (for each gene location i you wish to specify)

      base.mutation-probability.i

    Any settings for #3 override #2, and both override #1.

    The only parameter which can be specified this way in VectorSpecies is at present mutation-probability. However a number of parameters are specified this way in subclasses.

    Parameters

    base.genome-size
    int >= 1 or one of: geometric, uniform    		 (size of the genome, or if 'geometric' or 'uniform', the algorithm used to size the initial genome)
    base.chunk-size
    1 <= int <= genome-size (default=1)    		 (the chunk size for crossover (crossover will only occur on chunk boundaries))
    base.geometric-prob
    0.0 <= double < 1.0    		 (the coin-flip probability for increasing the initial size using the geometric distribution)
    base.min-initial-size
    int >= 0    		 (the minimum initial size of the genome)
    base.max-initial-size
    int >= min-initial-size    		 (the maximum initial size of the genome)
    base.crossover-type
    string, one of: one, two, any    		 (default crossover type (one-point, one-point-nonempty, two-point, two-point-nonempty, any-point (uniform), line, or intermediate)
    base.crossover-prob
    0.0 >= double >= 1.0     		(probability that a gene will get crossed over during any-point (uniform) or simulated binary crossover)
    base.line-extension
    double >= 0.0     		(for line and intermediate recombination, how far along the line or outside of the hypercube children can be. If this value is zero, all children must be within the hypercube.)
     
    base.mutation-prob   or
    base.segment.segment-number.mutation-prob   or
    base.mutation-prob.gene-number
    0.0 <= double <= 1.0     		(probability that a gene will get mutated over default mutation)

    Default Base
    vector.species

    See Also:
    Serialized Form

    • Field Detail

      • V_ONE_POINT_NO_NOP

        public static final java.lang.String V_ONE_POINT_NO_NOP
        See Also:
        Constant Field Values

      • V_TWO_POINT_NO_NOP

        public static final java.lang.String V_TWO_POINT_NO_NOP
        See Also:
        Constant Field Values

      • V_INTERMED_RECOMB

        public static final java.lang.String V_INTERMED_RECOMB
        See Also:
        Constant Field Values

      • V_SIMULATED_BINARY

        public static final java.lang.String V_SIMULATED_BINARY
        See Also:
        Constant Field Values

      • P_CROSSOVER_DISTRIBUTION_INDEX

        public static final java.lang.String P_CROSSOVER_DISTRIBUTION_INDEX
        See Also:
        Constant Field Values

      • P_GEOMETRIC_PROBABILITY

        public static final java.lang.String P_GEOMETRIC_PROBABILITY
        See Also:
        Constant Field Values

      • P_DUPLICATE_RETRIES

        public static final java.lang.String P_DUPLICATE_RETRIES
        See Also:
        Constant Field Values

      • crossoverProbability

        public double crossoverProbability
        Probability that a gene will cross over -- ONLY used in V_ANY_POINT crossover

      • crossoverType

        public int crossoverType
        What kind of crossover do we have?

      • genomeSize

        public int genomeSize
        How big of a genome should we create on initialization?

      • crossoverDistributionIndex

        public int crossoverDistributionIndex
        What should the SBX distribution index be?

      • genomeResizeAlgorithm

        public int genomeResizeAlgorithm
        How should we reset the genome?

      • minInitialSize

        public int minInitialSize
        What's the smallest legal genome?

      • maxInitialSize

        public int maxInitialSize
        What's the largest legal genome?

      • genomeIncreaseProbability

        public double genomeIncreaseProbability
        With what probability would our genome be at least 1 larger than it is now during initialization?

      • chunksize

        public int chunksize
        How big of chunks should we define for crossover?

      • lineDistance

        public double lineDistance
        How far along the long a child can be located for line or intermediate recombination

      • dynamicInitialSize

        public boolean dynamicInitialSize
        Was the initial size determined dynamically?

    • Constructor Detail

      • VectorSpecies

        public VectorSpecies()

    • Method Detail

      • mutationProbability

        public double mutationProbability(int gene)

      • duplicateRetries

        public int duplicateRetries(int gene)

      • defaultBase

        public Parameter defaultBase()
        Description copied from interface: Prototype
        Returns the default base for this prototype. This should generally be implemented by building off of the static base() method on the DefaultsForm object for the prototype's package. This should be callable during setup(...).

      • setup

        public void setup(EvolutionState state,
                  Parameter base)
        Description copied from class: Species
        The default version of setup(...) loads requested pipelines and calls setup(...) on them and normalizes their probabilities. If your individual prototype might need to know special things about the species (like parameters stored in it), then when you override this setup method, you'll need to set those parameters BEFORE you call super.setup(...), because the setup(...) code in Species sets up the prototype.
        Specified by:
        setup in interface Prototype
        Specified by:
        setup in interface Setup
        Overrides:
        setup in class Species
        See Also:
        Prototype.setup(EvolutionState,Parameter)

      • newIndividual

        public Individual newIndividual(EvolutionState state,
                                int thread)
        Description copied from class: Species
        Provides a brand-new individual to fill in a population. The default form simply calls clone(), creates a fitness, sets evaluated to false, and sets the species. If you need to make a more custom genotype (as is the case for GPSpecies, which requires a light rather than deep clone), you will need to override this method as you see fit.
        Overrides:
        newIndividual in class Species
Skip navigation links

DMelt 3.0 © DataMelt by jWork.ORG

You see the box below because you did not login.