NAME

AI::Perceptron - An implementation of a Perceptron

SYNOPSIS

use AI::Perceptron;

DESCRIPTION

This module is meant to show how a single node of a neural network works to beginners in the field.

The only mode of training the weights supported at this point in time is the Stochastic Approximation of the Gradient-Descent model.

CONSTRUCTOR

new( [%args] )

Creates a new perceptron with the following properties:

Inputs => number of inputs (scalar)
N      => learning rate    (scalar)
W      => array ref of weights (applied to the inputs)

The Number of elements in W must be equal to the number of inputs plus one. This is because W[0] is the Perceptron's threshold (so W[1] corresponds to the first input's weight).

Default values are: 1, 0.05, and [random], respectively.

METHODS

weights( [@W] )

Sets/gets the perceptron's weights. This is useful between training sessions to see if the weights are actually changing. Note again that W[0] is the Perceptron's threshold.

train( $n, $training_examples )

This uses the Stochastic Approximation of the Gradient-Descent model to adjust the perceptron's weights in such a way to achieve the desired outputs given in the training examples.

Note that this training method may undo previous trainings!

SEE ALSO

Statistics::LTU, the ASCII model contained in Perceptron.pm.

REFERENCES

Machine Learning, by Tom M. Mitchell

AUTHOR

Steve Purkis <spurkis@epn.nu>

COPYRIGHT

Copyright (c) 1999, 2000 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.