view release on metacpan or  search on metacpan

lib/Algorithm/Evolutionary/Utils.pm  view on Meta::CPAN

sub entropy {
  my $population = shift;
  my %frequencies;
  map( (defined $_->{'_fitness'})?$frequencies{$_->{'_fitness'}}++:1, @$population );
  my $entropy = 0;
  my $gente = scalar(@$population); # Population size
  for my $f ( keys %frequencies ) {
    my $this_freq = $frequencies{$f}/$gente;
    $entropy -= $this_freq*log( $this_freq );
  }
  return $entropy;
}

=head2 genotypic_entropy( $population)

Computes the entropy using the well known Shannon's formula:
L<http://en.wikipedia.org/wiki/Information_entropy> 'to avoid botching
highlighting; in this case we use chromosome frequencies instead of
fitness. 

=cut

sub genotypic_entropy {
  my $population = shift;
  my %frequencies;
  map( $frequencies{$_->{'_str'}}++, @$population );
  my $entropy = 0;
  my $gente = scalar(@$population); # Population size
  for my $f ( keys %frequencies ) {
    my $this_freq = $frequencies{$f}/$gente;
    $entropy -= $this_freq*log( $this_freq );
  }
  return $entropy;
}

=head2 hamming( $string_a, $string_b )

Computes the number of bit positions that are different among two strings, the well known Hamming distance.

=cut

sub hamming {
    my ($string_a, $string_b) = @_;
    return ( ( $string_a ^ $string_b ) =~ tr/\1//);
}

=head2 consensus( $population, $rough = 0 )

Consensus sequence representing the majority value for each bit;
returns the consensus binary string. If "rough", then the bit is set only if the 
difference is bigger than 0.2 (60/40 proportion). Otherwise, it is set to C<->

=cut

sub consensus {
  my $population = shift;
  my $rough = shift;
  my @frequencies;
  for ( @$population ) {
      for ( my $i = 0; $i < length($_->{'_str'}); $i ++ ) {
	  if ( !$frequencies[$i] ) {
	      $frequencies[$i]={ 0 => 0,
				 1 => 0};
	  }
	  $frequencies[$i]->{substr($_->{'_str'}, $i, 1)}++;
      }
  }
  my $consensus;
  for my $f ( @frequencies ) {
    if ( !$rough ) {
      if ( $f->{'0'} > $f->{'1'} ) {
	$consensus.='0';
      } else {
	$consensus.='1';
      }
    } else {
      my $total =  $f->{'0'} +  $f->{'1'};
      my $difference = (abs( $f->{'0'} - $f->{'1'} ))/$total;
      if ( $difference < 0.2 ) {
	$consensus .= '-';
      } else {
	if ( $f->{'0'} > $f->{'1'} ) {
	  $consensus.='0';
	} else {
	  $consensus.='1';
	}
      }
    }
  }
  return $consensus;
}

=head2 average( $population )

Computes an average of population fitness

=cut

sub average {
  my $population = shift;
  my @frequencies;
  my @fitnesses = map( $_->{'_fitness'}, @$population );
  return mean( @fitnesses );

}

=head2 random_bitstring( $bits[, $chromify = 0 ] )

Returns a random bitstring with the stated number of bits. Useful for
testing, mainly. Returns a "chromosome-like" object if $chromify is 1,
mainly putting it into a hashref with the C<_str> key.

=cut

sub random_bitstring {
  my $bits = shift || croak "No bits!";
  my $chromify = shift || 0;
  my $generator = new String::Random;
  my $regex = "\[01\]{$bits}";
  my $this_string = $generator->randregex($regex);
  return $chromify?{_str => $this_string}:$this_string;
}

=head2 random_number_array( $dimensions [, $min = -1] [, $range = 2] )

Returns a random number array with the stated length. Useful for testing, mainly.

=cut

sub random_number_array {
  my $dimensions = shift || croak "Null dimension!";
  my $min = shift || -1;
  my $range = shift || 2;

  my @array;
  for ( my $i = 0; $i < $dimensions; $i ++ ) {
    push @array, $min + rand($range);
  }
  return @array;
}


=head2 decode_string( $chromosome, $gene_size, $min, $range )

Decodes to a vector, each one of whose components ranges between $min
and $max. Returns that vector.

It does not work for $gene_size too big. Certainly not for 64, maybe for 32.

=cut

sub decode_string {
  my ( $chromosome, $gene_size, $min, $range ) = @_;

  my @output_vector;
  my $max_range = eval "0b"."1"x$gene_size;
  for (my $i = 0; $i < length($chromosome)/$gene_size; $i ++ ) {
    my $substr = substr( $chromosome, $i*$gene_size, $gene_size );
    push @output_vector, (($range - $min) * eval("0b$substr") / $max_range) + $min; 
  }
  return @output_vector;
}

=head2 vector_compare( $vector_1, $vector_2 )

Compares vectors, returns 1 if 1 dominates 2, -1 if it's the other way
round, and 0 if neither dominates the other. Both vectors are supposed
to be numeric. Returns C<0> if neither is bigger, and they are not
equal. Fails if the length is not the same. None of the combinations
above, returns C<undef>.

=cut

sub vector_compare {
  my ( $vector_1, $vector_2 ) = @_;

  if ( scalar @$vector_1 != scalar @$vector_2 ) {
    croak "Different lengths, can't compare\n";
  }

  my $length = scalar @$vector_1;
  my @results = map( $vector_1->[$_] <=> $vector_2->[$_], 0..($length-1));
  my %comparisons;
  map( $comparisons{$_}++, @results );
  if ( $comparisons{1} && !$comparisons{-1} ) {
    return 1;
  }
  if ( !$comparisons{1} && $comparisons{-1} ) {
    return -1;
  }
  if ( defined $comparisons{0} && $comparisons{0} == $length ) {
    return 0;
  }
  return undef;
}

=head1 SEE ALSO

This is a spin off from L<Algorithm::Evolutionary> so it's worth the while to check it out. And the spinning was due to finding I needed to include it in examples for the much simpler L<Algorithm::Evolutionary::Simple>. 

=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

=cut

"Still there?";