view release on metacpan or  search on metacpan

lib/Algorithm/Evolutionary/Op/EDA_step.pm  view on Meta::CPAN

    for ( 1..$population_size ) {
      my $indi = new Algorithm::Evolutionary::Individual::BitString $number_of_bits ; #Creates random individual
      $indi->evaluate( $onemax );
      push( @pop, $indi );
    }

    my $selector = new Algorithm::Evolutionary::Op::RouletteWheel $population_size; #One of the possible selectors

    my $generation = 
      new Algorithm::Evolutionary::Op::EDA_step( $onemax, $selector, $replacement_rate );

    my @sortPop = sort { $b->Fitness() <=> $a->Fitness() } @pop;
    my $bestIndi = $sortPop[0];
    my $previous_average = average( \@sortPop );
    $generation->apply( \@sortPop );

=head1 Base Class

L<Algorithm::Evolutionary::Op::Base>

=head1 DESCRIPTION

Estimation of Distribution Algorithms shun operators and instead try
to model the distribution of "good" solutions in the population. This
version corresponds to the most basic one.

=head1 METHODS

=cut

package Algorithm::Evolutionary::Op::EDA_step;

use lib qw(../../..);

our ($VERSION) = ( '$Revision: 1.5 $ ' =~ / (\d+\.\d+)/ ) ;

use Carp;

use base 'Algorithm::Evolutionary::Op::Base';

use Algorithm::Evolutionary qw(Hash_Wheel Individual::String);

# Class-wide constants
our $APPLIESTO =  'ARRAY';
our $ARITY = 1;

=head2 new( $evaluation_function, $replacement_rate )

Creates an algorithm, with no defaults except for the default
replacement operator (defaults to L<Algorithm::Evolutionary::Op::ReplaceWorst>)

=cut

sub new {
  my $class = shift;
  my $self = {};
  $self->{_eval} = shift || croak "No eval function found";
  $self->{_replacementRate} = shift || 0.5; #Default to half  replaced
  $self->{_population_size} = shift || 100; #Default
  $self->{_alphabet} = shift || [ 0, 1]; #Default
  bless $self, $class;
  return $self;
}


=head2 set( $ref_to_params_hash, $ref_to_code_hash, $ref_to_operators_hash )

Sets the instance variables. Takes a ref-to-hash as
input. Not intended to be used from outside the class

=cut

sub set {
  my $self = shift;
  my $hashref = shift || croak "No params here";
  my $codehash = shift || croak "No code here";
  my $opshash = shift || croak "No ops here";

  for ( keys %$codehash ) {
	$self->{"_$_"} =  eval "sub { $codehash->{$_} } ";
  }

  $self->{_ops} =();
  for ( keys %$opshash ) {
    push @{$self->{_ops}}, 
      Algorithm::Evolutionary::Op::Base::fromXML( $_, $opshash->{$_}->[1], $opshash->{$_}->[0] ) ;
  }
}

=head2 reset( $population )

Start all over again by resetting the population

=cut

sub reset {
  my $self = shift;
  my $population = shift;
  my $length = $population->[0]->size;
  @$population = ();
  my @alphabet = @{$self->{'_alphabet'}};
  for ( my $p= 0; $p < $self->{'_population_size'}; $p++ ) {
    my $string = '';
    for ( my $i = 0; $i < $length; $i++ ) {
      $string .= $alphabet[rand( @alphabet )];
    }
    my $new_one =  Algorithm::Evolutionary::Individual::String->fromString( $string );
    push @$population, $new_one;
  } 
}

=head2 apply( $population )

Applies the algorithm to the population, which should have
been evaluated first; checks that it receives a
ref-to-array as input, croaks if it does not. Returns a sorted,
culled, evaluated population for next generation.

=cut

sub apply ($) {