view release on metacpan or  search on metacpan

BackProp.pm  view on Meta::CPAN

		my $t0 		 =	new Benchmark;
        $self->{RUN}->run($map);
		$self->{LAST_TIME}=timestr(timediff(new Benchmark, $t0));
        return $self->map();
	}
    
    # This automatically uncrunches a response after running it
	sub run_uc {
    	$_[0]->uncrunch(run(@_));
    }

	# Returns benchmark and loop's ran or learned
	# for last run(), or learn()
	# operation preformed.
	#
	sub benchmarked {
		my $self	=	shift;
		return $self->{LAST_TIME};
	}
	    
	# Used to retrieve map from last internal run operation.
	sub map {
		my $self	 =	  shift;
		$self->{MAP}->map();
	}
	
	# Forces network to learn pattern passed and give desired
	# 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);
		$res  = $self->crunch($res)  if($res  == 0);
		
		# Fill in empty spaces at end of results matrix with a 0
		if($#{$res}<$out) {
			for my $x ($#{$res}+1..$out) {
				#$res->[$x] = 0;
			}
		}
		
		# Debug
		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);
			
			# Retrieve last mapping  and initialize a few variables.
			$map	=	$self->map();
			$y		=	$size-$div;
			$flag	=	1;
			
			# Compare the result map we just ran with the desired result map.
			$diff 	=	pdiff($map,$res);
			
			# This adjusts the increment multiplier to decrease as the loops increase
			if($_mi > $_mx) {
				$dinc *= 0.1;
				$_mi   = 0;
			} 
			
			$_mi++;
				
			
			# We de-increment the loop ammount to prevent infinite learning loops.
			# In old versions of this module, if you used too high of an initial input
			# $inc, then the network would keep jumping back and forth over your desired 
			# results because the increment was too high...it would try to push close to
			# the desired result, only to fly over the other edge too far, therby trying
			# to come back, over shooting again. 
			# This simply adjusts the learning gradient proportionally to the ammount of
			# convergance left as the difference decreases.
           	$inc   -= ($dinc*$diff);
			$inc   = 0.0000000001 if($inc < 0.0000000001);
			
			# This prevents it from seeming to get stuck in one location
			# by attempting to boost the values out of the hole they seem to be in.
			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)) {