Code fails without any reason?

Bluepixel has asked for the wisdom of the Perl Monks concerning the following question:

Dear Monks, why does the following code fail on line 137 with an out of memory error? Any help would be appreciated. Thanks

#!/usr/bin/perl

use strict;
use warnings;

sub sigm
    {
    my $x = shift;
    return (1/(1 + exp(-$x)));
    }

sub dxsigm
    {
    my $x = shift;
    return ($x*(1-$x));
    }


my $c_IN = 35;
my $c_HIDDEN =10;
my $c_OUT = 10;
my $c_EPSILON = 0.005;
my $c_NUMTRAIN = 10;

my @actafer;
$actafer[0]=[0,1,1,1,1,1,0,1,0,0,0,0,0,1,1,0,0,0,0,0,1,1,0,0,0,0,0,1,0
+,1,1,1,1,1,0];
$actafer[1]=[0,0,0,0,0,0,0,0,1,0,0,0,0,1,1,1,1,1,1,1,1,0,0,0,0,0,0,1,0
+,0,0,0,0,0,0];
$actafer[2]=[0,1,0,0,0,0,1,1,0,0,0,0,1,1,1,0,0,0,1,0,1,1,0,0,1,0,0,1,0
+,1,1,0,0,0,1];
$actafer[3]=[1,0,0,0,0,1,0,1,0,0,0,0,0,1,1,0,0,1,0,0,1,1,1,1,0,1,0,1,1
+,0,0,0,1,1,0];
$actafer[4]=[0,0,0,1,1,0,0,0,0,1,0,1,0,0,0,1,0,0,1,0,0,1,1,1,1,1,1,1,0
+,0,0,0,1,0,0];
$actafer[5]=[1,1,1,0,0,1,0,1,0,1,0,0,0,1,1,0,1,0,0,0,1,1,0,1,0,0,0,1,1
+,0,0,1,1,1,0];
$actafer[6]=[0,0,1,1,1,1,0,0,1,0,1,0,0,1,1,0,0,1,0,0,1,1,0,0,1,0,0,1,0
+,0,0,0,1,1,0];
$actafer[7]=[1,0,0,0,0,0,0,1,0,0,0,0,0,0,1,0,0,1,1,1,1,1,0,1,0,0,0,0,1
+,1,0,0,0,0,0];
$actafer[8]=[0,1,1,0,1,1,0,1,0,0,1,0,0,1,1,0,0,1,0,0,1,1,0,0,1,0,0,1,0
+,1,1,0,1,1,0];
$actafer[9]=[0,1,1,0,0,0,0,1,0,0,1,0,0,1,1,0,0,1,0,0,1,1,0,0,1,0,1,0,0
+,1,1,1,1,0,0];

my @desout;
$desout[0]=[1,0,0,0,0,0,0,0,0,0];
$desout[1]=[0,1,0,0,0,0,0,0,0,0];
$desout[2]=[0,0,1,0,0,0,0,0,0,0];
$desout[3]=[0,0,0,1,0,0,0,0,0,0];
$desout[4]=[0,0,0,0,1,0,0,0,0,0];
$desout[5]=[0,0,0,0,0,1,0,0,0,0];
$desout[6]=[0,0,0,0,0,0,1,0,0,0];
$desout[7]=[0,0,0,0,0,0,0,1,0,0];
$desout[8]=[0,0,0,0,0,0,0,0,1,0];
$desout[9]=[0,0,0,0,0,0,0,0,0,1];

my (@inhiddw,@hidoutw,@deltaihw,@deltahow,@x,@y,@z,@ehid,@eout,@ecm,@p
+atr,@matrizin);
my $delta=0.5;
my $alfa=0.1;


sub init
    {
    my ($i,$j);
   # srand48();
    for($i=0;$i<$c_IN;$i++) {
    for($j=0;$j<$c_HIDDEN;$j++) {
    $inhiddw[$i][$j] = -0.5 + rand;#;drand48();
    $deltaihw[$i][$j] = 0;
    }
    }

    for($i=0;$i<$c_HIDDEN;$i++)
       {
       for($j=0;$j<$c_OUT;$j++) {
         $hidoutw[$i][$j] = -0.5 + rand;#;drand48();
         $deltahow[$i][$j] = 0;
         }
       }

  for($i=0;$i<$c_NUMTRAIN;$i++)
   {
   $patr[$i] = 0;
   }

  for($i=0;$i<$c_IN;$i++)     #here was 35
    {
    $matrizin[$i]=0;
    }

    }

sub training
    {

 my ($i,$l,$num)=(0,0,0);
 my ($j,$t,$rep,$p);
 do {
   do {

   #/* select a random training pattern: i = (int)(NUMTRAIN*rnd), wher
+e 0<rnd<1 */
   #i = (int)(NUMTRAIN*(float) rand() / RAND_MAX);
   #$i = int($c_NUMTRAIN * rand()/10);
   $i = int (rand(10));
  } while($patr[$i]);

  for($rep=0;$rep<3;$rep++) {
   $j++;
   #my @array = @{$actafer[$i]};
   &netanswer($i);
   &backprop($i);
  }

  #print "J: $j\n";

  if(!($j%102)) #/*showerr();*/
   {
   printf("\n%ld",$j);
   }
  &error();

  $l = 1;
  for($t=0;$t<$c_NUMTRAIN;$t++) {
  #print "$ecm[$t]:$patr[$t]\n";
   $patr[$t] = $ecm[$t] < $c_EPSILON;
   $l = $l && ($patr[$t]);
  }
 } while(!$l);

printf("\n\n End of training\n");


    }

sub netanswer
    {
    my $in_i = shift;
    my ($i,$j,$totin);

    for($i=0;$i<$c_IN;$i++)
      {

      print "$i $c_IN $in_i\n";

      $x[$i] = $actafer[$in_i][$i];

      if (not defined $x[$i]){die;}

      }

    for($j=0;$j<$c_HIDDEN;$j++) {
      $totin = 0;
      for($i=0;$i<$c_IN;$i++)
      {
      #print "$x[$i] : $inhiddw[$i][$j]\n";
      $totin = $totin + $x[$i]*$inhiddw[$i][$j];
      }
    $y[$j] = &sigm($totin);
    }

 for($j=0;$j<$c_OUT;$j++) {
  $totin = 0;
  for($i=0;$i<$c_HIDDEN;$i++)
     {
     $totin = $totin + $y[$i]*$hidoutw[$i][$j];
     }
  $z[$j] = &sigm($totin);

   }


    }


sub backprop
    {
    my $k =shift;

    my ($i,$j,$temp);
    &betaout($k);
    &betahid();

    for($i=0;$i<$c_HIDDEN;$i++)
    {
    for($j=0;$j<$c_OUT;$j++) {
    $temp = -$delta*$y[$i]*$z[$j]*(1-$z[$j])*$eout[$j];
    $hidoutw[$i][$j] = $hidoutw[$i][$j] + $temp + $alfa*$deltahow[$i][
+$j];
    $deltahow[$i][$j] = $temp;
    }
    }

    for($i=0;$i<$c_IN;$i++)
      {
      for($j=0;$j<$c_HIDDEN;$j++) {
      $temp = -$delta*$x[$i]*$y[$j]*(1-$y[$j])*$ehid[$j];
      $inhiddw[$i][$j] = $inhiddw[$i][$j] + $temp + $alfa*$deltaihw[$i
+][$j];
      $deltaihw[$i][$j] = $temp;
        }
      }
    }

sub betaout
    {
    my $i = shift;
    my $j;
    for($j=0;$j<$c_OUT;$j++)
    {
    $eout[$j] = 0;
    }
    for($j=0;$j<$c_OUT;$j++)
    {
    #print "J: $j $i\n";
    $eout[$j] = $z[$j] - $desout[$i][$j];
    }

    }

sub betahid
    {
    my ($i,$j);
    for($i=0;$i<$c_HIDDEN;$i++)
       {$ehid[$i] = 0;
       }

    for($i=0;$i<$c_HIDDEN;$i++)
      {
      for($j=0;$j<$c_OUT;$j++)
         {
         $ehid[$i] = $ehid[$i] + $hidoutw[$i][$j]*$z[$j]*(1-$z[$j])*$e
+out[$j];
         }
      }
    }

sub error
    {
    my ($i);

    for($i=0;$i<$c_NUMTRAIN;$i++) {
    &netanswer($actafer[$i]);
   # $ecm[$i]=ec(\@{$desout[$i]},\@z,$c_OUT);
    $ecm[$i]=ec($i,$c_OUT);
    }


    }

sub ec
{
 my ($in_i,$SIZE) =@_;
 my ($i,$e) =(0,0);

 for($i=0;$i<$SIZE;$i++)
 {
 #print "I: $in_i $i\n";
 $e = $e + ($desout[$in_i][$i] - $z[$i])*($desout[$in_i][$i] - $z[$i])
+;
 }
 $e = 0.5 * $e;
 return $e;
 }


&main;

sub main
    {
    &init;
    &training;
   # &test;
    }
[download]

The original c code can be found here.

Comment on Code fails without any reason? Download Code

Replies are listed 'Best First'.
Re: Code fails without any reason? by Improv (Pilgrim) on May 01, 2003 at 15:04 UTC
I inserted debugging code. It's failing in the OTHER place you're calling netanswer: `sub error { my ($i); for($i=0;$i<$c_NUMTRAIN;$i++) { print "netanswer called from error()\n"; &netanswer($actafer[$i]); # $ecm[$i]=ec(\@{$desout[$i]},\@z,$c_OUT); $ecm[$i]=ec($i,$c_OUT); }` [download] There, you're just passing it $actafer[$i], which is an array reference instead of a value. The array reference then gets turned into an absurdly high index, and perl attempts to expand the array to that big, and fails. Fix error(), and the problem will go away.	[reply] [d/l]
Re: Re: Code fails without any reason? by Bluepixel (Beadle) on May 01, 2003 at 15:16 UTC
Thanks a lot. I must have overseen that second call :(. I thought the error was produced by the first call. I should have inserted a print command after the loop and probably would have found the error myself.	[reply]
Re: Code fails without any reason? by dragonchild (Archbishop) on May 01, 2003 at 14:51 UTC
I'm assuming you mean this line: `{ print "$i $c_IN $in_i\n"; # <----- $x[$i] = $actafer[$in_i][$i];` [download] I have no idea. Try reducing the amount of pre-initialization you do. Of course, that might require that you re-examine the algorithm and attempt to write it in Perl, not C-Perl. You're not taking advantage of Perl's capabilities, so it's going to waste (more) memory. However, unless you're working with a small box, you shouldn't run out of memory without some loops and push or unshift or splice to make your arrays big. shrugs In case you're wondering why this answer isn't that helpful ... I'm not going to parse and learn your code without you paying me. I'll help you after you explain what's going on. "Help!" doesn't cut it. ------ We are the carpenters and bricklayers of the Information Age. Don't go borrowing trouble. For programmers, this means Worry only about what you need to implement. Please remember that I'm crufty and crochety. All opinions are purely mine and all code is untested, unless otherwise specified.	[reply] [d/l]
Re: Re: Code fails without any reason? by Bluepixel (Beadle) on May 01, 2003 at 14:57 UTC
`for($i=0;$i<$c_IN;$i++) { print "$i $c_IN $in_i\n"; $x[$i] = $actafer[$in_i][$i]; if (not defined $x[$i]){die;} }` [download] The $i's are counter up to 34, as they should, then $i is reset to 0 (in the same loop) which causes `$x[$i] = $actafer[$in_i][$i];` to fail. Output is: `........ 28 35 9 29 35 9 30 35 9 31 35 9 32 35 9 33 35 9 34 35 9 0 35 ARRAY(0xa04116c) Out of memory during "large" request for 1073745920 bytes, total sbrk( +) is 31744 0 bytes at neuronalnetwork.pl line 137.` [download] I was wondering why $i is reset to 0?	[reply] [d/l] [select]
Re3: Code fails without any reason? by dragonchild (Archbishop) on May 01, 2003 at 15:15 UTC
Your problem may be related to $i being set to 0, but the actual issue is that $in_i is being assigned to an array reference. That is what's causing the memory issue. Check to see what you're doing in the error() function. :-) ------ We are the carpenters and bricklayers of the Information Age. Don't go borrowing trouble. For programmers, this means Worry only about what you need to implement. Please remember that I'm crufty and crochety. All opinions are purely mine and all code is untested, unless otherwise specified.	[reply]
Re: Code fails without any reason? by crenz (Priest) on May 01, 2003 at 21:35 UTC
As an aside, there's a couple interesting modules on CPAN: A number of implementations of different neural networks (including backpropagation networks) and bindings for libsvm (Support Vector Machines). They might save you a lot of time. I know what I'm talking about -- In one project I spent too much time implementing Neural Networks in Java when there's already dozens of implementations out there :).	[reply]