line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package AI::NeuralNet::SOM::Rect; |
2
|
|
|
|
|
|
|
|
3
|
3
|
|
|
3
|
|
47065
|
use strict; |
|
3
|
|
|
|
|
6
|
|
|
3
|
|
|
|
|
119
|
|
4
|
3
|
|
|
3
|
|
17
|
use warnings; |
|
3
|
|
|
|
|
5
|
|
|
3
|
|
|
|
|
128
|
|
5
|
|
|
|
|
|
|
|
6
|
3
|
|
|
3
|
|
1721
|
use Data::Dumper; |
|
3
|
|
|
|
|
17034
|
|
|
3
|
|
|
|
|
277
|
|
7
|
3
|
|
|
3
|
|
19
|
use base qw(AI::NeuralNet::SOM); |
|
3
|
|
|
|
|
6
|
|
|
3
|
|
|
|
|
1713
|
|
8
|
3
|
|
|
3
|
|
2896
|
use AI::NeuralNet::SOM::Utils; |
|
3
|
|
|
|
|
7
|
|
|
3
|
|
|
|
|
5158
|
|
9
|
|
|
|
|
|
|
|
10
|
|
|
|
|
|
|
=pod |
11
|
|
|
|
|
|
|
|
12
|
|
|
|
|
|
|
=head1 NAME |
13
|
|
|
|
|
|
|
|
14
|
|
|
|
|
|
|
AI::NeuralNet::SOM::Rect - Perl extension for Kohonen Maps (rectangular topology) |
15
|
|
|
|
|
|
|
|
16
|
|
|
|
|
|
|
=head1 SYNOPSIS |
17
|
|
|
|
|
|
|
|
18
|
|
|
|
|
|
|
use AI::NeuralNet::SOM::Rect; |
19
|
|
|
|
|
|
|
my $nn = new AI::NeuralNet::SOM::Rect (output_dim => "5x6", |
20
|
|
|
|
|
|
|
input_dim => 3); |
21
|
|
|
|
|
|
|
$nn->initialize; |
22
|
|
|
|
|
|
|
$nn->train (30, |
23
|
|
|
|
|
|
|
[ 3, 2, 4 ], |
24
|
|
|
|
|
|
|
[ -1, -1, -1 ], |
25
|
|
|
|
|
|
|
[ 0, 4, -3]); |
26
|
|
|
|
|
|
|
|
27
|
|
|
|
|
|
|
print $nn->as_data; |
28
|
|
|
|
|
|
|
|
29
|
|
|
|
|
|
|
=head1 INTERFACE |
30
|
|
|
|
|
|
|
|
31
|
|
|
|
|
|
|
=head2 Constructor |
32
|
|
|
|
|
|
|
|
33
|
|
|
|
|
|
|
The constructor takes the following arguments (additionally to those in the base class): |
34
|
|
|
|
|
|
|
|
35
|
|
|
|
|
|
|
=over |
36
|
|
|
|
|
|
|
|
37
|
|
|
|
|
|
|
=item C : (mandatory, no default) |
38
|
|
|
|
|
|
|
|
39
|
|
|
|
|
|
|
A string of the form "3x4" defining the X and the Y dimensions. |
40
|
|
|
|
|
|
|
|
41
|
|
|
|
|
|
|
=back |
42
|
|
|
|
|
|
|
|
43
|
|
|
|
|
|
|
Example: |
44
|
|
|
|
|
|
|
|
45
|
|
|
|
|
|
|
my $nn = new AI::NeuralNet::SOM::Rect (output_dim => "5x6", |
46
|
|
|
|
|
|
|
input_dim => 3); |
47
|
|
|
|
|
|
|
|
48
|
|
|
|
|
|
|
=cut |
49
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
sub new { |
51
|
8
|
|
|
8
|
0
|
2831
|
my $class = shift; |
52
|
8
|
|
|
|
|
34
|
my %options = @_; |
53
|
8
|
|
|
|
|
41
|
my $self = bless { %options }, $class; |
54
|
|
|
|
|
|
|
|
55
|
8
|
50
|
|
|
|
71
|
if ($self->{output_dim} =~ /(\d+)x(\d+)/) { |
56
|
8
|
50
|
|
|
|
57
|
$self->{_X} = $1 and $self->{_Y} = $2; |
57
|
|
|
|
|
|
|
} else { |
58
|
0
|
|
|
|
|
0
|
die "output dimension does not have format MxN"; |
59
|
|
|
|
|
|
|
} |
60
|
8
|
50
|
|
|
|
20
|
if ($self->{input_dim} > 0) { |
61
|
8
|
|
|
|
|
23
|
$self->{_Z} = $self->{input_dim}; |
62
|
|
|
|
|
|
|
} else { |
63
|
0
|
|
|
|
|
0
|
die "input dimension must be positive integer"; |
64
|
|
|
|
|
|
|
} |
65
|
|
|
|
|
|
|
|
66
|
|
|
|
|
|
|
|
67
|
8
|
|
|
|
|
40
|
($self->{_R}) = map { $_ / 2 } sort {$b <= $a } ($self->{_X}, $self->{_Y}); # radius |
|
16
|
|
|
|
|
44
|
|
|
8
|
|
|
|
|
43
|
|
68
|
8
|
|
33
|
|
|
53
|
$self->{_Sigma0} = $options{sigma0} || $self->{_R}; # impact distance, start value |
69
|
8
|
|
50
|
|
|
52
|
$self->{_L0} = $options{learning_rate} || 0.1; # learning rate, start value |
70
|
8
|
|
|
|
|
29
|
return $self; |
71
|
|
|
|
|
|
|
} |
72
|
|
|
|
|
|
|
|
73
|
|
|
|
|
|
|
=pod |
74
|
|
|
|
|
|
|
|
75
|
|
|
|
|
|
|
=head2 Methods |
76
|
|
|
|
|
|
|
|
77
|
|
|
|
|
|
|
=cut |
78
|
|
|
|
|
|
|
|
79
|
|
|
|
|
|
|
sub initialize { |
80
|
4
|
|
|
4
|
1
|
25
|
my $self = shift; |
81
|
4
|
|
|
|
|
8
|
my @data = @_; |
82
|
|
|
|
|
|
|
|
83
|
4
|
|
|
|
|
8
|
our $i = 0; |
84
|
|
|
|
|
|
|
my $get_from_stream = sub { |
85
|
0
|
0
|
|
0
|
|
0
|
$i = 0 if $i > $#data; |
86
|
0
|
|
|
|
|
0
|
return [ @{ $data[$i++] } ]; # cloning ! |
|
0
|
|
|
|
|
0
|
|
87
|
4
|
50
|
|
|
|
34
|
} if @data; |
88
|
|
|
|
|
|
|
$get_from_stream ||= sub { |
89
|
120
|
|
|
120
|
|
169
|
return [ map { rand( 1 ) - 0.5 } 1..$self->{_Z} ]; |
|
360
|
|
|
|
|
878
|
|
90
|
4
|
|
100
|
|
|
40
|
}; |
91
|
|
|
|
|
|
|
|
92
|
4
|
|
|
|
|
15
|
for my $x (0 .. $self->{_X}-1) { |
93
|
20
|
|
|
|
|
35
|
for my $y (0 .. $self->{_Y}-1) { |
94
|
120
|
|
|
|
|
174
|
$self->{map}->[$x]->[$y] = &$get_from_stream; |
95
|
|
|
|
|
|
|
} |
96
|
|
|
|
|
|
|
} |
97
|
|
|
|
|
|
|
} |
98
|
|
|
|
|
|
|
|
99
|
|
|
|
|
|
|
sub bmu { |
100
|
8823
|
|
|
8823
|
1
|
11338
|
my $self = shift; |
101
|
8823
|
|
|
|
|
9203
|
my $sample = shift; |
102
|
|
|
|
|
|
|
|
103
|
8823
|
|
|
|
|
16183
|
my $closest; # [x,y, distance] value and co-ords of closest match |
104
|
8823
|
|
|
|
|
18273
|
for my $x (0 .. $self->{_X}-1) { |
105
|
44115
|
|
|
|
|
117915
|
for my $y (0 .. $self->{_Y}-1){ |
106
|
264690
|
|
|
|
|
792365
|
my $distance = AI::NeuralNet::SOM::Utils::vector_distance ($self->{map}->[$x]->[$y], $sample); # || Vi - Sample || |
107
|
|
|
|
|
|
|
#warn "distance to $x, $y : $distance"; |
108
|
264690
|
100
|
|
|
|
572325
|
$closest = [0, 0, $distance] unless $closest; |
109
|
264690
|
100
|
|
|
|
819904
|
$closest = [$x, $y, $distance] if $distance < $closest->[2]; |
110
|
|
|
|
|
|
|
} |
111
|
|
|
|
|
|
|
} |
112
|
8823
|
|
|
|
|
40228
|
return @$closest; |
113
|
|
|
|
|
|
|
} |
114
|
|
|
|
|
|
|
|
115
|
|
|
|
|
|
|
|
116
|
|
|
|
|
|
|
sub neighbors { # http://www.ai-junkie.com/ann/som/som3.html |
117
|
7380
|
|
|
7380
|
1
|
9800
|
my $self = shift; |
118
|
7380
|
|
|
|
|
8122
|
my $sigma = shift; |
119
|
7380
|
|
|
|
|
7710
|
my $X = shift; |
120
|
7380
|
|
|
|
|
7575
|
my $Y = shift; |
121
|
|
|
|
|
|
|
|
122
|
7380
|
|
|
|
|
7216
|
my @neighbors; |
123
|
7380
|
|
|
|
|
17574
|
for my $x (0 .. $self->{_X}-1) { |
124
|
36900
|
|
|
|
|
76893
|
for my $y (0 .. $self->{_Y}-1){ |
125
|
221400
|
|
|
|
|
388221
|
my $distance = sqrt ( ($x - $X) * ($x - $X) + ($y - $Y) * ($y - $Y) ); |
126
|
221400
|
100
|
|
|
|
468259
|
next if $distance > $sigma; |
127
|
45998
|
|
|
|
|
137299
|
push @neighbors, [ $x, $y, $distance ]; # we keep the distances |
128
|
|
|
|
|
|
|
} |
129
|
|
|
|
|
|
|
} |
130
|
7380
|
|
|
|
|
25118
|
return \@neighbors; |
131
|
|
|
|
|
|
|
} |
132
|
|
|
|
|
|
|
|
133
|
|
|
|
|
|
|
=pod |
134
|
|
|
|
|
|
|
|
135
|
|
|
|
|
|
|
=cut |
136
|
|
|
|
|
|
|
|
137
|
|
|
|
|
|
|
sub radius { |
138
|
2
|
|
|
2
|
1
|
3266
|
my $self = shift; |
139
|
2
|
|
|
|
|
10
|
return $self->{_R}; |
140
|
|
|
|
|
|
|
} |
141
|
|
|
|
|
|
|
|
142
|
|
|
|
|
|
|
=pod |
143
|
|
|
|
|
|
|
|
144
|
|
|
|
|
|
|
=over |
145
|
|
|
|
|
|
|
|
146
|
|
|
|
|
|
|
=item I |
147
|
|
|
|
|
|
|
|
148
|
|
|
|
|
|
|
I<$m> = I<$nn>->map |
149
|
|
|
|
|
|
|
|
150
|
|
|
|
|
|
|
This method returns the 2-dimensional array of vectors in the grid (as a reference to an array of |
151
|
|
|
|
|
|
|
references to arrays of vectors). The representation of the 2-dimensional array is straightforward. |
152
|
|
|
|
|
|
|
|
153
|
|
|
|
|
|
|
Example: |
154
|
|
|
|
|
|
|
|
155
|
|
|
|
|
|
|
my $m = $nn->map; |
156
|
|
|
|
|
|
|
for my $x (0 .. 5) { |
157
|
|
|
|
|
|
|
for my $y (0 .. 4){ |
158
|
|
|
|
|
|
|
warn "vector at $x, $y: ". Dumper $m->[$x]->[$y]; |
159
|
|
|
|
|
|
|
} |
160
|
|
|
|
|
|
|
} |
161
|
|
|
|
|
|
|
|
162
|
|
|
|
|
|
|
=cut |
163
|
|
|
|
|
|
|
|
164
|
|
|
|
|
|
|
sub as_string { |
165
|
2
|
|
|
2
|
1
|
1070
|
my $self = shift; |
166
|
2
|
|
|
|
|
5
|
my $s = ''; |
167
|
|
|
|
|
|
|
|
168
|
2
|
|
|
|
|
7
|
$s .= " "; |
169
|
2
|
|
|
|
|
10
|
for my $y (0 .. $self->{_Y}-1){ |
170
|
12
|
|
|
|
|
40
|
$s .= sprintf (" %02d ",$y); |
171
|
|
|
|
|
|
|
} |
172
|
2
|
|
|
|
|
6
|
$s .= sprintf "\n","-"x107,"\n"; |
173
|
|
|
|
|
|
|
|
174
|
2
|
|
|
|
|
4
|
my $dim = scalar @{ $self->{map}->[0]->[0] }; |
|
2
|
|
|
|
|
10
|
|
175
|
|
|
|
|
|
|
|
176
|
2
|
|
|
|
|
9
|
for my $x (0 .. $self->{_X}-1) { |
177
|
10
|
|
|
|
|
48
|
for my $w ( 0 .. $dim-1 ){ |
178
|
30
|
|
|
|
|
56
|
$s .= sprintf ("%02d | ",$x); |
179
|
30
|
|
|
|
|
64
|
for my $y (0 .. $self->{_Y}-1){ |
180
|
180
|
|
|
|
|
716
|
$s .= sprintf ("% 2.2f ", $self->{map}->[$x]->[$y]->[$w]); |
181
|
|
|
|
|
|
|
} |
182
|
30
|
|
|
|
|
58
|
$s .= sprintf "\n"; |
183
|
|
|
|
|
|
|
} |
184
|
10
|
|
|
|
|
20
|
$s .= sprintf "\n"; |
185
|
|
|
|
|
|
|
} |
186
|
2
|
|
|
|
|
22
|
return $s; |
187
|
|
|
|
|
|
|
} |
188
|
|
|
|
|
|
|
|
189
|
|
|
|
|
|
|
=pod |
190
|
|
|
|
|
|
|
|
191
|
|
|
|
|
|
|
=item I |
192
|
|
|
|
|
|
|
|
193
|
|
|
|
|
|
|
print I<$nn>->as_data |
194
|
|
|
|
|
|
|
|
195
|
|
|
|
|
|
|
This methods creates a string containing the raw vector data, row by |
196
|
|
|
|
|
|
|
row. This can be fed into gnuplot, for instance. |
197
|
|
|
|
|
|
|
|
198
|
|
|
|
|
|
|
=cut |
199
|
|
|
|
|
|
|
|
200
|
|
|
|
|
|
|
sub as_data { |
201
|
2
|
|
|
2
|
1
|
4
|
my $self = shift; |
202
|
2
|
|
|
|
|
6
|
my $s = ''; |
203
|
|
|
|
|
|
|
|
204
|
2
|
|
|
|
|
5
|
my $dim = scalar @{ $self->{map}->[0]->[0] }; |
|
2
|
|
|
|
|
7
|
|
205
|
2
|
|
|
|
|
11
|
for my $x (0 .. $self->{_X}-1) { |
206
|
10
|
|
|
|
|
18
|
for my $y (0 .. $self->{_Y}-1){ |
207
|
60
|
|
|
|
|
73
|
for my $w ( 0 .. $dim-1 ){ |
208
|
180
|
|
|
|
|
526
|
$s .= sprintf ("\t%f", $self->{map}->[$x]->[$y]->[$w]); |
209
|
|
|
|
|
|
|
} |
210
|
60
|
|
|
|
|
83
|
$s .= sprintf "\n"; |
211
|
|
|
|
|
|
|
} |
212
|
|
|
|
|
|
|
} |
213
|
2
|
|
|
|
|
22
|
return $s; |
214
|
|
|
|
|
|
|
} |
215
|
|
|
|
|
|
|
|
216
|
|
|
|
|
|
|
=pod |
217
|
|
|
|
|
|
|
|
218
|
|
|
|
|
|
|
=back |
219
|
|
|
|
|
|
|
|
220
|
|
|
|
|
|
|
|
221
|
|
|
|
|
|
|
=head1 SEE ALSO |
222
|
|
|
|
|
|
|
|
223
|
|
|
|
|
|
|
L |
224
|
|
|
|
|
|
|
|
225
|
|
|
|
|
|
|
=head1 AUTHOR |
226
|
|
|
|
|
|
|
|
227
|
|
|
|
|
|
|
Robert Barta, Erho@devc.atE |
228
|
|
|
|
|
|
|
|
229
|
|
|
|
|
|
|
=head1 COPYRIGHT AND LICENSE |
230
|
|
|
|
|
|
|
|
231
|
|
|
|
|
|
|
Copyright (C) 2007 by Robert Barta |
232
|
|
|
|
|
|
|
|
233
|
|
|
|
|
|
|
This library is free software; you can redistribute it and/or modify |
234
|
|
|
|
|
|
|
it under the same terms as Perl itself, either Perl version 5.8.8 or, |
235
|
|
|
|
|
|
|
at your option, any later version of Perl 5 you may have available. |
236
|
|
|
|
|
|
|
|
237
|
|
|
|
|
|
|
=cut |
238
|
|
|
|
|
|
|
|
239
|
|
|
|
|
|
|
our $VERSION = '0.02'; |
240
|
|
|
|
|
|
|
|
241
|
|
|
|
|
|
|
1; |
242
|
|
|
|
|
|
|
|
243
|
|
|
|
|
|
|
__END__ |