AI-NeuralNet-BackProp
view release on metacpan or search on metacpan
BackProp.pm view on Meta::CPAN
$self->{OUTPUTS}->{SIZE} = ++$oid;
return $self->{OUTPUTS}->{LIST}->[$oid]->{ID};
}
1;
package AI::NeuralNet::BackProp;
use Benchmark;
use strict;
# Returns the number of elements in an array ref, undef on error
sub _FETCHSIZE {
my $a=$_[0];
my ($b,$x);
return undef if(substr($a,0,5) ne "ARRAY");
foreach $b (@{$a}) { $x++ };
return $x;
}
# Debugging subs
$AI::NeuralNet::BackProp::DEBUG = 0;
BackProp.pm view on Meta::CPAN
#
# It returns a percentage string.
#
sub learn_set {
my $self = shift if(substr($_[0],0,4) eq 'AI::');
my $data = shift;
my %args = @_;
my $len = $#{$data}/2-1;
my $inc = $args{inc};
my $max = $args{max};
my $error = $args{error};
my $p = (defined $args{flag}) ?$args{flag} :1;
my $row = (defined $args{pattern})?$args{pattern}*2+1:1;
my ($fa,$fb);
for my $x (0..$len) {
print "\nLearning index $x...\n" if($AI::NeuralNet::BackProp::DEBUG);
my $str = $self->learn( $data->[$x*2], # The list of data to input to the net
$data->[$x*2+1], # The output desired
inc=>$inc, # The starting learning gradient
max=>$max, # The maximum num of loops allowed
error=>$error); # The maximum (%) error allowed
print $str if($AI::NeuralNet::BackProp::DEBUG);
}
my $res;
$data->[$row] = $self->crunch($data->[$row]) if($data->[$row] == 0);
if ($p) {
$res=pdiff($data->[$row],$self->run($data->[$row-1]));
} else {
BackProp.pm view on Meta::CPAN
#
# It returns a true value.
#
sub learn_set_rand {
my $self = shift if(substr($_[0],0,4) eq 'AI::');
my $data = shift;
my %args = @_;
my $len = $#{$data}/2-1;
my $inc = $args{inc};
my $max = $args{max};
my $error = $args{error};
my @learned;
while(1) {
_GET_X:
my $x=$self->intr(rand()*$len);
goto _GET_X if($learned[$x]);
$learned[$x]=1;
print "\nLearning index $x...\n" if($AI::NeuralNet::BackProp::DEBUG);
my $str = $self->learn($data->[$x*2], # The list of data to input to the net
$data->[$x*2+1], # The output desired
inc=>$inc, # The starting learning gradient
max=>$max, # The maximum num of loops allowed
error=>$error); # The maximum (%) error allowed
print $str if($AI::NeuralNet::BackProp::DEBUG);
}
return 1;
}
# Returns the index of the element in array REF passed with the highest comparative value
sub high {
shift if(substr($_[0],0,4) eq 'AI::');
BackProp.pm view on Meta::CPAN
# results. See usage in POD.
sub learn {
my $self = shift;
my $omap = shift;
my $res = shift;
my %args = @_;
my $inc = $args{inc} || 0.20;
my $max = $args{max} || 1024;
my $_mx = intr($max/10);
my $_mi = 0;
my $error = ($args{error}>-1 && defined $args{error}) ? $args{error} : -1;
my $div = $self->{DIV};
my $size = $self->{SIZE};
my $out = $self->{OUT};
my $divide = AI::NeuralNet::BackProp->intr($div/$out);
my ($a,$b,$y,$flag,$map,$loop,$diff,$pattern,$value);
my ($t0,$it0);
no strict 'refs';
# Take care of crunching strings passed
$omap = $self->crunch($omap) if($omap == 0);
BackProp.pm view on Meta::CPAN
AI::NeuralNet::BackProp::out1 "Num output neurons: $out, Input neurons: $size, Division: $divide\n";
# Start benchmark timer and initalize a few variables
$t0 = new Benchmark;
$flag = 0;
$loop = 0;
my $ldiff = 0;
my $dinc = 0.0001;
my $cdiff = 0;
$diff = 100;
$error = ($error>-1)?$error:-1;
# $flag only goes high when all neurons in output map compare exactly with
# desired result map or $max loops is reached
#
while(!$flag && ($max ? $loop<$max : 1)) {
$it0 = new Benchmark;
# Run the map
$self->{RUN}->run($omap);
BackProp.pm view on Meta::CPAN
if($diff eq $ldiff) {
$cdiff++;
$inc += ($dinc*$diff)+($dinc*$cdiff*10);
} else {
$cdiff=0;
}
# Save last $diff
$ldiff = $diff;
# This catches a max error argument and handles it
if(!($error>-1 ? $diff>$error : 1)) {
$flag=1;
last;
}
# Debugging
AI::NeuralNet::BackProp::out4 "Difference: $diff\%\t Increment: $inc\tMax Error: $error\%\n";
AI::NeuralNet::BackProp::out1 "\n\nMapping results from $map:\n";
# This loop compares each element of the output map with the desired result map.
# If they don't match exactly, we call weight() on the offending output neuron
# and tell it what it should be aiming for, and then the offending neuron will
# try to adjust the weights of its synapses to get closer to the desired output.
# See comments in the weight() method of AI::NeuralNet::BackProp for how this works.
my $l=$self->{NET};
for my $i (0..$out-1) {
$a = $map->[$i];
BackProp.pm view on Meta::CPAN
UPDATED: You can now learn inputs with a 0 value. Beware though, it may not learn() a 0 value
in the input map if you have randomness disabled. See NOTES on using a 0 value with randomness
disabled.
The first two arguments may be array refs (or now, strings), and they may be of different lengths.
Options should be written on hash form. There are three options:
inc => $learning_gradient
max => $maximum_iterations
error => $maximum_allowable_percentage_of_error
$learning_gradient is an optional value used to adjust the weights of the internal
connections. If $learning_gradient is ommitted, it defaults to 0.20.
$maximum_iterations is the maximum numbers of iteration the loop should do.
It defaults to 1024. Set it to 0 if you never want the loop to quit before
the pattern is perfectly learned.
$maximum_allowable_percentage_of_error is the maximum allowable error to have. If
this is set, then learn() will return when the perecentage difference between the
actual results and desired results falls below $maximum_allowable_percentage_of_error.
If you do not include 'error', or $maximum_allowable_percentage_of_error is set to -1,
then learn() will not return until it gets an exact match for the desired result OR it
reaches $maximum_iterations.
=item $net->learn_set(\@set, [ options ]);
UPDATE: Inputs and outputs in the dataset can now be strings. See information on auto-crunching
in learn()
This takes the same options as learn() and allows you to specify a set to learn, rather
BackProp.pm view on Meta::CPAN
=head1 NOTES
=item Learning 0s With Randomness Disabled
You can now use 0 values in any input maps. This is a good improvement over versions 0.40
and 0.42, where no 0s were allowed because the learning would never finish learning completly
with a 0 in the input.
Yet with the allowance of 0s, it requires one of two factors to learn correctly. Either you
must enable randomness with $net->random(0.0001) (Any values work [other than 0], see random() ),
or you must set an error-minimum with the 'error => 5' option (you can use some other error value
as well).
When randomness is enabled (that is, when you call random() with a value other than 0), it interjects
a bit of randomness into the output of every neuron in the network, except for the input and output
neurons. The randomness is interjected with rand()*$rand, where $rand is the value that was
passed to random() call. This assures the network that it will never have a pure 0 internally. It is
bad to have a pure 0 internally because the weights cannot change a 0 when multiplied by a 0, the
product stays a 0. Yet when a weight is multiplied by 0.00001, eventually with enough weight, it will
be able to learn. With a 0 value instead of 0.00001 or whatever, then it would never be able
to add enough weight to get anything other than a 0.
The second option to allow for 0s is to enable a maximum error with the 'error' option in
learn() , learn_set() , and learn_set_rand() . This allows the network to not worry about
learning an output perfectly.
For accuracy reasons, it is recomended that you work with 0s using the random() method.
If anyone has any thoughts/arguments/suggestions for using 0s in the network, let me know
at jdb@wcoil.com.
Now I know you people are out there that are using the module...
I can hear the fists hitting the keyboards in frustration. :-) Relieve
some of that frustration by e-mailing me and letting me know what
you think of the module and any suggestions you got. Especially you
guys in HP Labs and at Xerox! :-)
Use it, let me know what you all think. This is just a
groud-up write of a neural network, no code stolen or
anything else. It uses the -IDEA- of back-propagation
for error correction, with the -IDEA- of the delta
rule and hopefield theory, as I understand them. So, don't expect
a classicist view of nerual networking here. I simply wrote
from operating theory, not math theory. Any die-hard neural
networking gurus out there? Let me know how far off I am with
this code! :-)
Regards,
~ Josiah Bryan, <jdb@wcoil.com>
<P>Note, the old method of calling crunch on the values still works just as well.</P>
<P><B>UPDATED:</B> You can now learn inputs with a 0 value. Beware though, it may not <A HREF="#item_learn"><CODE>learn()</CODE></A> a 0 value
in the input map if you have randomness disabled. See NOTES on using a 0 value with randomness
disabled.</P>
<P>The first two arguments may be array refs (or now, strings), and they may be of different lengths.</P>
<P>Options should be written on hash form. There are three options:
</P>
<PRE>
inc => $learning_gradient
max => $maximum_iterations
error => $maximum_allowable_percentage_of_error</PRE>
<P>$learning_gradient is an optional value used to adjust the weights of the internal
connections. If $learning_gradient is ommitted, it defaults to 0.20.
</P>
<P>
$maximum_iterations is the maximum numbers of iteration the loop should do.
It defaults to 1024. Set it to 0 if you never want the loop to quit before
the pattern is perfectly learned.</P>
<P>$maximum_allowable_percentage_of_error is the maximum allowable error to have. If
this is set, then <A HREF="#item_learn"><CODE>learn()</CODE></A> will return when the perecentage difference between the
actual results and desired results falls below $maximum_allowable_percentage_of_error.
If you do not include 'error', or $maximum_allowable_percentage_of_error is set to -1,
then <A HREF="#item_learn"><CODE>learn()</CODE></A> will not return until it gets an exact match for the desired result OR it
reaches $maximum_iterations.</P>
<P></P>
<DT><STRONG><A NAME="item_learn_set">$net->learn_set(\@set, [ options ]);</A></STRONG><BR>
<DD>
<B>UPDATED:</B> Inputs and outputs in the dataset can now be strings. See information on auto-crunching
in <A HREF="#item_learn"><CODE>learn()</CODE></A>
<P>This takes the same options as <A HREF="#item_learn"><CODE>learn()</CODE></A> and allows you to specify a set to learn, rather
than individual patterns. A dataset is an array refrence with at least two elements in the
array, each element being another array refrence (or now, a scalar string). For each pattern to
<HR SIZE=1 COLOR=BLACK>
<H1><A NAME="notes">NOTES</A></H1>
<DL>
<DT><STRONG><A NAME="item_Learning_0s_With_Randomness_Disabled">Learning 0s With Randomness Disabled</A></STRONG><BR>
<DD>
You can now use 0 values in any input maps. This is a good improvement over versions 0.40
and 0.42, where no 0s were allowed because the learning would never finish learning completly
with a 0 in the input.
<P>Yet with the allowance of 0s, it requires one of two factors to learn correctly. Either you
must enable randomness with $net-><A HREF="#item_random"><CODE>random(0.0001)</CODE></A> (Any values work [other than 0], see <A HREF="#item_random"><CODE>random()</CODE></A> ),
or you must set an error-minimum with the 'error => 5' option (you can use some other error value
as well).</P>
<P>When randomness is enabled (that is, when you call <A HREF="#item_random"><CODE>random()</CODE></A> with a value other than 0), it interjects
a bit of randomness into the output of every neuron in the network, except for the input and output
neurons. The randomness is interjected with rand()*$rand, where $rand is the value that was
passed to <A HREF="#item_random"><CODE>random()</CODE></A> call. This assures the network that it will never have a pure 0 internally. It is
bad to have a pure 0 internally because the weights cannot change a 0 when multiplied by a 0, the
product stays a 0. Yet when a weight is multiplied by 0.00001, eventually with enough weight, it will
be able to learn. With a 0 value instead of 0.00001 or whatever, then it would never be able
to add enough weight to get anything other than a 0.</P>
<P>The second option to allow for 0s is to enable a maximum error with the 'error' option in
<A HREF="#item_learn"><CODE>learn()</CODE></A> , <A HREF="#item_learn_set"><CODE>learn_set()</CODE></A> , and <A HREF="#item_learn_set_rand"><CODE>learn_set_rand()</CODE></A> . This allows the network to not worry about
learning an output perfectly.</P>
<P>For accuracy reasons, it is recomended that you work with 0s using the <A HREF="#item_random"><CODE>random()</CODE></A> method.</P>
<P>If anyone has any thoughts/arguments/suggestions for using 0s in the network, let me know
at <A HREF="mailto:jdb@wcoil.com.">jdb@wcoil.com.</A></P>
<P></P></DL>
<P>
<HR SIZE=1 COLOR=BLACK>
<H1><A NAME="other included packages">OTHER INCLUDED PACKAGES</A></H1>
<DL>
<P>Neural Nets are formed by simulated neurons connected together much the same
way the brain's neurons are, neural networks are able to associate and
generalize without rules. They have solved problems in pattern recognition,
robotics, speech processing, financial predicting and signal processing, to
name a few.</P>
<P>One of the first impressive neural networks was NetTalk, which read in ASCII
text and correctly pronounced the words (producing phonemes which drove a
speech chip), even those it had never seen before. Designed by John Hopkins
biophysicist Terry Sejnowski and Charles Rosenberg of Princeton in 1986,
this application made the Backprogagation training algorithm famous. Using
the same paradigm, a neural network has been trained to classify sonar
returns from an undersea mine and rock. This classifier, designed by
Sejnowski and R. Paul Gorman, performed better than a nearest-neighbor
classifier.</P>
<P>The kinds of problems best solved by neural networks are those that people
are good at such as association, evaluation and pattern recognition.
Problems that are difficult to compute and do not require perfect answers,
just very good answers, are also best done with neural networks. A quick,
designers are working on weather forecasts by neural networks (Myself
included). Currently, doctors are developing medical neural networks as an
aid in diagnosis. Attorneys and insurance companies are also working on
neural networks to help estimate the value of claims.</P>
<P>Neural networks are poor at precise calculations and serial processing. They
are also unable to predict or recognize anything that does not inherently
contain some sort of pattern. For example, they cannot predict the lottery,
since this is a random process. It is unlikely that a neural network could
be built which has the capacity to think as well as a person does for two
reasons. Neural networks are terrible at deduction, or logical thinking and
the human brain is just too complex to completely simulate. Also, some
problems are too difficult for present technology. Real vision, for
example, is a long way off.</P>
<P>In short, Neural Networks are poor at precise calculations, but good at
association, evaluation, and pattern recognition.
</P>
<P>
<HR SIZE=1 COLOR=BLACK>
<A HREF="http://www.josiah.countystart.com/modules/AI/rec.pl?docs.htm">AI::NeuralNet::BackProp</a> - <i>Written by Josiah Bryan, <<A HREF="mailto:jdb@wcoil.com">jdb@wcoil.com</A>></I>
examples/ex_add2.pl view on Meta::CPAN
my @list;
my $t1=new Benchmark;
for my $a (1..$top)
{
print "Outer Loop: $a : ";
$forgetfulness = $net->learn_set( \@data,
inc => $inc,
max => 500,
error => -1);
print "Forgetfulness: $forgetfulness %\n";
}
my $t2=new Benchmark;
$runtime = timediff($t2,$t1);
print "run took ",timestr($runtime),"\n";
examples/ex_bmp.pl view on Meta::CPAN
# Make it learn the whole dataset $top times
my @list;
my $top=3;
for my $a (0..$top) {
my $t1=new Benchmark;
print "\n\nOuter Loop: $a\n";
# Test fogetfullness
my $f = $net->learn_set(\@data, inc => 0.1,
max => 500,
error => -1);
# Print it
print "\n\nForgetfullness: $f%\n";
# Save net to disk
$net->save('images.net');
my $t2=new Benchmark;
my $td=timediff($t2,$t1);
print "\nLoop $a took ",timestr($td),"\n";
examples/ex_dow.pl view on Meta::CPAN
# Make it learn the whole dataset $top times
my @list;
my $top=1;
for my $a (0..$top) {
my $t1=new Benchmark;
print "\n\nOuter Loop: $a\n";
# Test fogetfullness
my $f = $net->learn_set(\@data, inc => 0.2,
max => 2000,
error => -1);
# Print it
print "\n\nForgetfullness: $f%\n";
# Save net to disk
$net->save('dow.dat');
my $t2=new Benchmark;
my $td=timediff($t2,$t1);
print "\nLoop $a took ",timestr($td),"\n";
( run in 1.197 second using v1.01-cache-2.11-cpan-49f99fa48dc )