http://qs321.pair.com?node_id=355979

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

Hey all.... playing with AI for the first time...So I got AI::Neuralnet::backprop to play with. I thought how about trying to get the thing to identify even numbers? Well I failed miserably ... I attached my code and the sample output it produced. It got numbers wrong that I explicitly gave it in its pattern set to learn. I don't understand the problem with the code and was trying to learn how the neural net thingy works so I'm a bit mystified. Am I missing something? Any help or suggestions welcome.

CODE

use AI::NeuralNet::BackProp; my $net = new AI::NeuralNet::BackProp(2,5,1); # Add a small amount of randomness to the network $net->random(0.001); $net->range ([0,1]); $net->debug (4); # Create a data set to learn my @set = ( [ 0,0,0,0,0 ], [ 1 ], [ 0,0,0,1,0 ], [ 1 ], [ 0,0,1,0,0 ], [ 1 ], [ 0,0,1,1,0 ], [ 1 ], [ 0,0,0,0,1 ], [ 0 ], [ 0,0,0,1,1 ], [ 0 ], [ 0,0,1,0,1 ], [ 0 ], [ 0,0,1,1,1 ], [ 0 ], ); my $f = $net->learn_set (\@set); print "Forgetfulness: $f unit\n"; # Run a test phrase through the network for ( $i = 0 ; $i < 17 ;$i++){ @bit= (0,0,0,0,0); #reset bits $bit[0] = 1 if ( ($i & 16) == 16 ); $bit[1] = 1 if ( ($i & 8) == 8 ); $bit[2] = 1 if ( ($i & 4) == 4 ); $bit[3] = 1 if ( ($i & 2) == 2 ); $bit[4] = 1 if ( ($i & 1) == 1 ); print "Testing $i @bit -> "; my $result = $net->run( \@bit); my $answer = $$result[0]; print "$answer\n"; }
RESULTS
Forgetfulness: 0.0000000000 Testing 0 0 0 0 0 0 -> 1 Testing 1 0 0 0 0 1 -> 1 Testing 2 0 0 0 1 0 -> 1 Testing 3 0 0 0 1 1 -> 0 Testing 4 0 0 1 0 0 -> 1 Testing 5 0 0 1 0 1 -> 0 Testing 6 0 0 1 1 0 -> 0 Testing 7 0 0 1 1 1 -> 0 Testing 8 0 1 0 0 0 -> 1 Testing 9 0 1 0 0 1 -> 0 Testing 10 0 1 0 1 0 -> 0 Testing 11 0 1 0 1 1 -> 0 Testing 12 0 1 1 0 0 -> 0 Testing 13 0 1 1 0 1 -> 0 Testing 14 0 1 1 1 0 -> 0 Testing 15 0 1 1 1 1 -> 0 Testing 16 1 0 0 0 0 -> 1