NAME
    AI::Perceptron - example of a node in a neural network.

SYNOPSIS
     use AI::Perceptron;

     my $p = AI::Perceptron->new
               ->num_inputs( 2 )
               ->learning_rate( 0.04 )
               ->threshold( 0.02 )
               ->weights([ 0.1, 0.2 ]);

     my @inputs  = ( 1.3, -0.45 );   # input can be any number
     my $target  = 1;                # output is always -1 or 1
     my $current = $p->compute_output( @inputs );

     print "current output: $current, target: $target\n";

     $p->add_examples( [ $target, @inputs ] );

     $p->max_iterations( 10 )->train or
       warn "couldn't train in 10 iterations!";

     print "training until it gets it right\n";
     $p->max_iterations( -1 )->train; # watch out for infinite loops

DESCRIPTION
    This module is meant to show how a single node of a neural network
    works.

    Training is done by the *Stochastic Approximation of the
    Gradient-Descent* model.

MODEL
    Model of a Perceptron

                  +---------------+
     X[1] o------ |W[1]      T    |
     X[2] o------ |W[2] +---------+         +-------------------+
      .           | .   |   ___   |_________|    __  Squarewave |_______\  Output
      .           | .   |   \     |    S    | __|    Generator  |       /
      .           | .   |   /__   |         +-------------------+
     X[n] o------ |W[n] |   Sum   |
                  +-----+---------+

                 S  =  T + Sum( W[i]*X[i] )  as i goes from 1 -> n
            Output  =  1 if S > 0; else -1

    Where "X[n]" are the perceptron's *inputs*, "W[n]" are the *Weights*
    that get applied to the corresponding input, and "T" is the *Threshold*.

    The *squarewave generator* just turns the result into a positive or
    negative number.

    So in summary, when you feed the perceptron some numeric inputs you get
    either a positive or negative output depending on the input's weights
    and a threshold.

TRAINING
    Usually you have to train a perceptron before it will give you the
    outputs you expect. This is done by giving the perceptron a set of
    examples containing the output you want for some given inputs:

        -1 => -1, -1
        -1 =>  1, -1
        -1 => -1,  1
         1 =>  1,  1

    If you've ever studied boolean logic, you should recognize that as the
    truth table for an "AND" gate (ok so we're using -1 instead of the
    commonly used 0, same thing really).

    You *train* the perceptron by iterating over the examples and adjusting
    the *weights* and *threshold* by some value until the perceptron's
    output matches the expected output of each example:

        while some examples are incorrectly classified
            update weights for each example that fails

    The value each weight is adjusted by is calculated as follows:

        delta[i] = learning_rate * (expected_output - output) * input[i]

    Which is know as a negative feedback loop - it uses the current output
    as an input to determine what the next output will be.

    Also, note that this means you can get stuck in an infinite loop. It's
    not a bad idea to set the maximum number of iterations to prevent that.

CONSTRUCTOR
    new( [%args] )
        Creates a new perceptron with the following default properties:

            num_inputs    = 1
            learning_rate = 0.01
            threshold     = 0.0
            weights       = empty list

        Ideally you should use the accessors to set the properties, but for
        backwards compatability you can still use the following arguments:

            Inputs => $number_of_inputs  (positive int)
            N      => $learning_rate     (float)
            W      => [ @weights ]       (floats)

        The number of elements in *W* must be equal to the number of inputs
        plus one. This is because older version of AI::Perceptron combined
        the threshold and the weights a single list where W[0] was the
        threshold and W[1] was the first weight. Great idea, eh? :) That's
        why it's *DEPRECATED*.

ACCESSORS
    num_inputs( [ $int ] )
        Set/get the perceptron's number of inputs.

    learning_rate( [ $float ] )
        Set/get the perceptron's number of inputs.

    weights( [ \@weights ] )
        Set/get the perceptron's weights (floats).

        For backwards compatability, returns a list containing the
        *threshold* as the first element in list context:

          ($threshold, @weights) = $p->weights;

        This usage is *DEPRECATED*.

    threshold( [ $float ] )
        Set/get the perceptron's number of inputs.

    training_examples( [ \@examples ] )
        Set/get the perceptron's list of training examples. This should be a
        list of arrayrefs of the form:

            [ $expected_result => @inputs ]

    max_iterations( [ $int ] )
        Set/get the perceptron's number of inputs, a negative value implies
        no maximum.

METHODS
    compute_output( @inputs )
        Computes and returns the perceptron's output (either -1 or 1) for
        the given inputs. See the above model for more details.

    add_examples( @training_examples )
        Adds the @training_examples to to current list of examples. See
        training_examples() for more details.

    train( [ @training_examples ] )
        Uses the *Stochastic Approximation of the Gradient-Descent* model to
        adjust the perceptron's weights until all training examples are
        classified correctly.

        @training_examples can be passed for convenience. These are passed
        to add_examples(). If you want to re-train the perceptron with an
        entirely new set of examples, reset the training_examples().

AUTHOR
    Steve Purkis <spurkis@epn.nu>

COPYRIGHT
    Copyright (c) 1999-2003 Steve Purkis. All rights reserved.

    This package is free software; you can redistribute it and/or modify it
    under the same terms as Perl itself.

REFERENCES
    *Machine Learning*, by Tom M. Mitchell.

THANKS
    Himanshu Garg <himanshu@gdit.iiit.net> for his bug-report and feedback.
    Many others for their feedback.

SEE ALSO
    Statistics::LTU, AI::jNeural, AI::NeuralNet::BackProp,
    AI::NeuralNet::Kohonen