File Coverage

blib/lib/PDL/Opt/Simplex.pm
Criterion Covered Total %
statement 79 91 86.8
branch 17 26 65.3
condition 2 3 66.6
subroutine 5 5 100.0
pod 0 1 0.0
total 103 126 81.7


line stmt bran cond sub pod time code
1              
2             =head1 NAME
3              
4             PDL::Opt::Simplex -- Simplex optimization routines
5              
6             =head1 SYNOPSIS
7              
8             use PDL::Opt::Simplex;
9              
10             ($optimum,$ssize,$optval) = simplex($init,$initsize,$minsize,
11             $maxiter,
12             sub {evaluate_func_at($_[0])},
13             sub {display_simplex($_[0])}
14             );
15              
16             =head1 DESCRIPTION
17              
18             This package implements the commonly used simplex optimization
19             algorithm. The basic idea of the algorithm is to move
20             a "simplex" of N+1 points in the N-dimensional search space
21             according to certain rules. The main
22             benefit of the algorithm is that you do not need to calculate
23             the derivatives of your function.
24              
25             $init is a 1D vector holding the initial values of the N fitted
26             parameters, $optimum is a vector holding the final solution.
27             $optval is the evaluation of the final solution.
28              
29             $initsize is the size of $init (more...)
30              
31             $minsize is some sort of convergence criterion (more...)
32             - e.g. $minsize = 1e-6
33              
34             The sub is assumed to understand more than 1 dimensions and threading.
35             Its signature is 'inp(nparams); [ret]out()'. An example would be
36              
37             sub evaluate_func_at {
38             my($xv) = @_;
39             my $x1 = $xv->slice("(0)");
40             my $x2 = $xv->slice("(1)");
41             return $x1**4 + ($x2-5)**4 + $x1*$x2;
42             }
43              
44             Here $xv is a vector holding the current values of the parameters
45             being fitted which are then sliced out explicitly as $x1 and $x2.
46              
47             $ssize gives a very very approximate estimate of how close we might
48             be - it might be miles wrong. It is the euclidean distance between
49             the best and the worst vertices. If it is not very small, the algorithm
50             has not converged.
51              
52             =head1 FUNCTIONS
53              
54             =head2 simplex
55              
56             =for ref
57              
58             Simplex optimization routine
59              
60             =for usage
61              
62             ($optimum,$ssize,$optval) = simplex($init,$initsize,$minsize,
63             $maxiter,
64             sub {evaluate_func_at($_[0])},
65             sub {display_simplex($_[0])}
66             );
67              
68             See module C for more information.
69              
70             =head1 CAVEATS
71              
72             Do not use the simplex method if your function has local minima.
73             It will not work. Use genetic algorithms or simulated annealing
74             or conjugate gradient or momentum gradient descent.
75              
76             They will not really work either but they are not guaranteed not to work ;)
77             (if you have infinite time, simulated annealing is guaranteed to work
78             but only after it has visited every point in your space).
79              
80             =head1 SEE ALSO
81              
82             Ron Shaffer's chemometrics web page and references therein:
83             C.
84              
85             Numerical Recipes (bla bla bla XXX ref).
86              
87             The demonstration (Examples/Simplex/tsimp.pl and tsimp2.pl).
88              
89             =head1 AUTHOR
90              
91             Copyright(C) 1997 Tuomas J. Lukka.
92             All rights reserved. There is no warranty. You are allowed
93             to redistribute this software / documentation under certain
94             conditions. For details, see the file COPYING in the PDL
95             distribution. If this file is separated from the PDL distribution,
96             the copyright notice should be included in the file.
97              
98              
99              
100             =cut
101              
102             package PDL::Opt::Simplex;
103 1     1   67161 use PDL;
  1         3  
  1         9  
104 1     1   6 use PDL::Primitive;
  1         2  
  1         7  
105 1     1   8 use strict;
  1         2  
  1         21  
106 1     1   5 use PDL::Exporter;
  1         2  
  1         3  
107              
108             # use AutoLoader;
109              
110             @PDL::Opt::Simplex::ISA = qw/PDL::Exporter/;
111              
112             @PDL::Opt::Simplex::EXPORT_OK = qw/simplex/;
113             %PDL::Opt::Simplex::EXPORT_TAGS = ( Func => [@PDL::Opt::Simplex::EXPORT_OK] );
114              
115             *simplex = \&PDL::simplex;
116              
117             sub PDL::simplex {
118 2     2 0 8 my ( $init, $initsize, $minsize, $maxiter, $sub, $logsub, $t ) = @_;
119 2 50       8 if ( !defined $t ) { $t = 0 }
  2         6  
120 2         5 my ( $i, $j );
121 2         9 my ( $nd, $nd2 ) = ( dims($init), 1 );
122 2         5 my $simp;
123 2 100       10 if ( $nd2 == 1 ) {
    50          
124 1         6 $simp = PDL->zeroes( $nd, $nd + 1 );
125 1         4 $simp .= $init;
126              
127             # Constructing a tetrahedron:
128             # At step n (starting from zero)
129             # take vertices 0..n and move them 1/(n+1) to negative dir on axis n.
130             # Take vertex n+1 and move it n/(n+1) to positive dir on axis n
131 1 50       5 if ( !ref $initsize ) {
132 0         0 $initsize = PDL->pdl($initsize)->dummy( 0, $nd );
133             }
134 1         7 for ( $i = 0 ; $i < $nd ; $i++ ) {
135 2         18 my $pj = $i / ( $i + 1 );
136 2         14 ( my $stoopid = $simp->slice("$i,0:$i") ) -=
137             $initsize->at($i) * $pj;
138 2         13 ( my $stoopid1 = $simp->slice( "$i," . ( $i + 1 ) ) ) +=
139             $initsize->at($i) * ( 1 - $pj );
140             }
141             }
142             elsif ( $nd2 == $nd + 1 ) {
143 1         3 $simp = $init;
144             }
145             else {
146 0         0 return;
147             }
148 2         9 my $maxind = PDL->zeroes(2);
149 2         10 my $minind = PDL->null;
150 2         8 my $ssum = PDL->null;
151 2         6 my $worst;
152             my $new;
153 2         4 my $vals = &{$sub}($simp);
  2         6  
154 2         251 my $ss1 = ( $simp - $simp->slice(":,0") )**2;
155 2         56 sumover( $ss1, ( my $ss2 = PDL->null ) );
156 2         11 my $ssize = PDL::max( sqrt($ss2) );
157 2 50       13 &{$logsub}( $simp, $vals, $ssize )
  0         0  
158             if $logsub;
159              
160 2   66     19 while ( $maxiter-- and max( PDL->topdl($ssize) ) > $minsize ) {
161 101         297 my $valsn = $vals;
162 101 50       244 if ($t) {
163 0         0 my $noise = $vals->random();
164 0         0 $noise->random;
165 0         0 $valsn = $vals + $t * ( -log( $noise + 0.00001 ) );
166             }
167 101         1403 maximum_n_ind( $valsn, $maxind );
168 101         1148 minimum_ind( $valsn, $minind );
169 101         287 my @worstvals = map { $valsn->at( $maxind->at($_) ) } 0 .. 1;
  202         453  
170 101         258 my $bestval = $valsn->at($minind);
171              
172 101         1610 sumover( $simp->xchg( 0, 1 ), $ssum );
173 101         625 $ssum -= ( $worst = $simp->slice( ":,(" . $maxind->at(0) . ")" ) );
174 101         270 $ssum /= $nd;
175 101         2331 $new = 2 * $ssum - $worst;
176 101         544 my $val = ( &{$sub}($new) )->at(0);
  101         296  
177 101 50       638 if ($t) {
178 0         0 $val = $val - $t * ( -log( rand() + 0.00001 ) );
179             }
180 101         154 my $removetop = 0;
181 101 100       306 if ( $val < $bestval ) {
    100          
182 30         556 my $newnew = $new + $ssum - $worst;
183 30         128 my $val2 = &{$sub}($newnew);
  30         87  
184 30 100       1353 if ( $val2->at(0) < $val ) {
185             # print "CASE1 Reflection and Expansion\n";
186 22         57 $worst .= $newnew;
187 22         38 $val = $val2;
188             }
189             else {
190             # print "CASE2 Reflection, $newnew, $val, $val2\n";
191 8         26 $worst .= $new;
192             }
193 30         123 $removetop = 1;
194             }
195             elsif ( $val < $worstvals[1] ) {
196             # print "CASE3 Reflection\n";
197 12         33 $worst .= $new;
198 12         23 $removetop = 1;
199             }
200             else {
201 59         1794 my $newnew = 0.5 * $ssum + 0.5 * $worst;
202 59         496 my $val2 = &{$sub}($newnew);
  59         166  
203 59 50       2694 if ( $val2->at(0) < $worstvals[0] ) {
204             # print "CASE4 Contraction, $newnew, $val, $val2\n";
205 59         155 $worst .= $newnew;
206 59         114 $val = $val2;
207 59         229 $removetop = 1;
208             }
209             }
210 101 50       204 if ($removetop) {
211 101         224 ( my $stoopid = $vals->slice( "(" . $maxind->at(0) . ")" ) ) .= $val;
212             }
213             else {
214             # print "CASE5 Multiple Contraction\n";
215 0         0 $simp = 0.5 * $simp->slice(":,$minind") + 0.5 * $simp;
216 0         0 my $idx = which( sequence($nd+1) != $minind );
217 0         0 ( my $stoopid = $vals->index($idx) ) .= &{$sub}($simp->dice_axis(1,$idx));
  0         0  
218             }
219 101         559 my $ss1 = ( $simp - $simp->slice(":,0") )**2;
220 101         985 sumover( $ss1, ( $ss2 = PDL->null ) );
221 101         323 $ssize = PDL::max( sqrt($ss2) );
222 101 50       950 &{$logsub}( $simp, $vals, $ssize )
  0         0  
223             if $logsub;
224             }
225 2         10 minimum_ind( $vals, ( my $mmind = PDL->null ) );
226 2         9 return ( $simp->slice(":,$mmind"), $ssize, $vals->index($mmind) );
227             }
228              
229             1;