File Coverage

lib/Algorithm/Evolutionary/Wheel.pm

Criterion	Covered	Total	%
statement	50	50	100.0
branch	5	6	83.3
condition	3	5	60.0
subroutine	6	6	100.0
pod	3	3	100.0
total	67	70	95.7

use strict;
use warnings;

=head1 NAME

Algorithm::Evolutionary::Wheel - Random selector of things depending on probabilities

=head1 SYNOPSIS

    my $wheel = new Algorithm::Evolutionary::Wheel( @probs );
    print $wheel->spin(); #Returns an element according to probabilities;

=head1 DESCRIPTION

Creates a "roulette wheel" for spinning and selecting stuff. It will
be used in several places; mainly in the
L.  Take care that fitness
must be non-zero positives; since if they aren't, roulette wheel won't
work at all

=head1 METHODS

=cut

package Algorithm::Evolutionary::Wheel;
use Carp;

our ($VERSION) = ( '$Revision: 3.6 $ ' =~ / (\d+\.\d+)/ ) ;

=head2 new( @probabilites )

Creates a new roulette wheel. Takes an array of numbers, which need not be
normalized

=cut

sub new {
  my $class = shift;
  my @probs = @_;

  my $self;
  $self->{'_accProbs'} = [ 0 ];

  my $acc = 0;
  for ( @probs ) { $acc += $_;}
  croak "The sum of fitness is 0, can't use roulette wheel\n" if ! $acc;
  for ( @probs ) { $_ /= $acc;} #Normalizes array

  #Now creates the accumulated array
  my $aux = 0;
  for ( @probs ) {
	push @{$self->{'_accProbs'}}, $_ + $aux;
	$aux += $_;
  }
  bless $self, $class;
  return $self;
}

=head2 spin( [$number_of_individuals = 1])

Returns an individual whose probability is related to its fitness

=cut

sub spin {
  my $self = shift;
  my $number_of_individuals = shift || 1;
  my $i = 0;
  my @rand;
  for my $n ( 1..$number_of_individuals ) {
    push @rand, rand();
  }
  my @individuals;
  for ( my $r=0; $r<= $#rand; $r++ ) {
    my $i = first( $rand[$r], $self->{'_accProbs'} );
      # my $i = -1; # First iteration must be 0
      # do {
      # 	  $i++;
      # } until (( $acc_probs[$i+1] > $rand[$r] ) || ($i >= $#acc_probs ));
      # $individuals[$r] = $i;
    push @individuals, $i;
  }
  if ( $number_of_individuals > 1 ) {
    return @individuals;
  } else {
    return $individuals[0];
  }

}

=head2 first( $item, $ref_to_list )

Returns the index of the first individual smaller than the item

=cut

sub first {
  my $item = shift;
  my $list = shift || croak "No list";
  my $first=0;
  my $last= scalar @$list -1;
  my $mid=int($last/2);
  while ($first <= $last ) {
    if ( $item > $list->[$mid] ) {
      $first = $mid + 1;
    } else {
      $last = $mid -1;
    }
    $mid = $first+ int(($last - $first )/2);
  }
  return $last;

}
=head1 Copyright

  This file is released under the GPL. See the LICENSE file included in this distribution,
  or go to http://www.fsf.org/licenses/gpl.txt

  CVS Info: $Date: 2010/12/08 09:31:24 $
  $Header: /media/Backup/Repos/opeal/opeal/Algorithm-Evolutionary/lib/Algorithm/Evolutionary/Wheel.pm,v 3.6 2010/12/08 09:31:24 jmerelo Exp $
  $Author: jmerelo $

=cut

"The truth is by here";

#Test code
#my @array = qw( 5 4 3 2 1 );
#my $wheel = new Wheel @array;

#my @histo;
#for ( 0..100 ){
#  my $s = $wheel->spin();
#  print "$s\n";
#  $histo[$s]++;
#}

#for ( 0..(@histo - 1)){
#  print $_, " => $histo[$_] \n";
#}

#my @array2 = qw( 1 3 7 4 2 1 );
#my $wheel2 = new Wheel @array2;

#my @histo2;
#for ( 0..100 ){
#  my $s = $wheel2->spin();
#  print "$s\n";
#  $histo2[$s]++;
#}

#for ( 0..(@histo2 - 1)){
#  print $_, " => $histo2[$_] \n";
#}