blib/lib/Algorithm/Evolutionary/Op/Easy_MO.pm untested

filter: Statementstmt 0.0% 0 / 80 Branchbran 0.0% 0 / 2 Conditioncond 0.0% 0 / 15 MC/DCmcdc 0.0% 0 / 8 Subroutinesub 0.0% 0 / 9 Podpod 0.0% 0 / 3 totaltotal 0.0% 0 / 117 CC 2 SCAR 10.3
CC
2
Cov
0%
new
6.0
set
2.0
apply
2.0
CRAP 2.8
1 use strict; #-*-cperl-*-
2 use warnings;
3  
4 use lib qw( ../../.. );
5  
6 =head1 NAME
7  
8 Algorithm::Evolutionary::Op::Easy_MO - Multiobjecttive evolutionary algorithm, single generation, with
9 variable operators
10
11  
12 =head1 SYNOPSIS
13  
14 #Mutation and crossover. Default selection rate is 0.4
15 my $algo = new Algorithm::Evolutionary::Op::Easy_MO( $eval );
16  
17 #Define an easy single-generation algorithm with predefined mutation and crossover
18 my $m = new Algorithm::Evolutionary::Op::Bitflip; #Changes a single bit
19 my $c = new Algorithm::Evolutionary::Op::Crossover; #Classical 2-point crossover
20 my $generation = new Algorithm::Evolutionary::Op::Easy_MO( $rr, 0.2, [$m, $c] );
21  
22 =head1 Base Class
23  
24 L<Algorithm::Evolutionary::Op::Base>
25  
26 =cut
27  
28 =head1 DESCRIPTION
29  
30 "Easy" to use, single generation of an evolutionary algorithm. Takes
31 an arrayref of operators as input, or defines bitflip-mutation and
32 2-point crossover as default. The C<apply> method applies a single
33 iteration of the algorithm to the population it takes as input
34  
35 =head1 METHODS
36  
37 =cut
38  
39 package Algorithm::Evolutionary::Op::Easy_MO;
40  
41 our ($VERSION) = ( '$Revision: 3.6 $ ' =~ / (\d+\.\d+)/ ) ;
42  
43 use Carp;
44  
45 use Algorithm::Evolutionary qw( Wheel Op::Bitflip
46 Op::Crossover
47 Op::Eval::MO_Rank );
48  
49 use base 'Algorithm::Evolutionary::Op::Base';
50  
51 # Class-wide constants
52 our $APPLIESTO = 'ARRAY';
53  
54 =head2 new( $eval_func, [$selection_rate,] [$operators_arrayref] )
55  
56 Creates an algorithm that optimizes the handled fitness function and
57 reference to an array of operators. If this reference is null, an
58 array consisting of bitflip mutation and 2 point crossover is
59 generated. Which, of course, might not what you need in case you don't
60 have a binary chromosome. Take into account that in this case the
61 fitness function should return a reference to array.
62  
63 =cut
64  
65 sub new {
66 my $class = shift;
67 my $self = {};
68 $self->{_eval} = shift || croak "No eval function found";
69 $self->{_rank} = new Algorithm::Evolutionary::Op::Eval::MO_Rank $self->{'_eval'};
70 $self->{_selrate} = shift || 0.4;
71 if ( @_ ) {
72 $self->{_ops} = shift;
73 } else {
74 #Create mutation and crossover
75 my $mutation = new Algorithm::Evolutionary::Op::Bitflip;
76 push( @{$self->{_ops}}, $mutation );
77 my $xover = new Algorithm::Evolutionary::Op::Crossover;
78 push( @{$self->{_ops}}, $xover );
79 }
80 bless $self, $class;
81 return $self;
82  
83 }
84  
85 =head2 set( $hashref, codehash, opshash )
86  
87 Sets the instance variables. Takes a ref-to-hash (for options), codehash (for fitness) and opshash (for operators)
88  
89 =cut
90  
91 sub set {
92 my $self = shift;
93 my $hashref = shift || croak "No params here";
94 my $codehash = shift || croak "No code here";
95 my $opshash = shift || croak "No ops here";
96 $self->{_selrate} = $hashref->{selrate};
97  
98 for ( keys %$codehash ) {
99 $self->{"_$_"} = eval "sub { $codehash->{$_} } " || carp "Error compiling fitness function: $! => $@";
100 }
101  
102 $self->{_ops} =();
103 for ( keys %$opshash ) {
104 #First element of the array contains the content, second the rate.
105 push @{$self->{_ops}},
106 Algorithm::Evolutionary::Op::Base::fromXML( $_, $opshash->{$_}->[1], $opshash->{$_}->[0] );
107 }
108  
109 }
110  
111 =head2 apply( $population )
112  
113 Applies the algorithm to the population; checks that it receives a
114 ref-to-array as input, croaks if it does not. Returns a sorted,
115 culled, evaluated population for next generation.
116  
117 =cut
118  
119 sub apply ($) {
120 my $self = shift;
121 my $pop = shift || croak "No population here";
122  
123 #Evaluate
124 my $eval = $self->{_eval};
125 my @ops = @{$self->{_ops}};
126 $self->{'_rank'}->apply( $pop );
127  
128 #Sort
129 my @popsort = sort { $b->{_fitness} <=> $a->{_fitness}; } @$pop;
130  
131 #Cull
132 my $pringaos = int(($#popsort+1)*$self->{_selrate}); #+1 gives you size
133 # print "Pringaos $pringaos\n";
134 splice @popsort, - $pringaos;
135 # print "Población ", scalar @popsort, "\n";
136
137 #Reproduce
138 my @rates = map( $_->{'rate'}, @ops );
139 my $opWheel = new Algorithm::Evolutionary::Wheel @rates;
140  
141 #Generate offpring;
142 my $originalSize = $#popsort; # Just for random choice
143 for ( my $i = 0; $i < $pringaos; $i ++ ) {
144 my @offspring;
145 my $selectedOp = $ops[ $opWheel->spin()];
146 for ( my $j = 0; $j < $selectedOp->arity(); $j ++ ) {
147 my $chosen = $popsort[ int ( rand( $originalSize ) )];
148 push( @offspring, $chosen ); #No need to clone, it's not changed in ops
149 }
150 my $mutante = $selectedOp->apply( @offspring );
151 push( @popsort, $mutante );
152 }
153  
154 #Return
155 for ( my $i = 0; $i <= $#popsort; $i++ ) {
156 # print $i, "->", $popsort[$i]->asString, "\n";
157 $pop->[$i] = $popsort[$i];
158 }
159  
160
161 }
162  
163 =head1 SEE ALSO
164  
165 L<Algorithm::Evolutionary::Op::CanonicalGA>.
166 L<Algorithm::Evolutionary::Op::FullAlgorithm>.
167 L<Algorithm::Evolutionary::Op::Easy> for the scalar version of this code.
168  
169 =head1 Copyright
170
171 This file is released under the GPL. See the LICENSE file included in this distribution,
172 or go to http://www.fsf.org/licenses/gpl.txt
173  
174 CVS Info: $Date: 2011/02/14 06:55:36 $
175 $Header: /media/Backup/Repos/opeal/opeal/Algorithm-Evolutionary/lib/Algorithm/Evolutionary/Op/Easy_MO.pm,v 3.6 2011/02/14 06:55:36 jmerelo Exp $
176 $Author: jmerelo $
177 $Revision: 3.6 $
178 $Name $
179  
180 =cut
181  
182 "The truth is out there";