Implementation of sUpervised Classifier System (UCS)
---------------------------------------------------

Written by Narayanan U. Edakunni as part of the ADEPT project

The directory consists of matlab scripts that implements UCS. The UCS system
consists of input made up of bit strings. The classifier system itself consists
of condition-class tuple. The prediction is based on a set of rules selected
based on the input bit string. The rules of the population is evolved using
a GA and the fitness of the rules is computed based on the training error on
the training data set presented to the classifier system in an online fashion.

Here the condition of the rules is represented as a ternary string with the 
# symbol being represented by the digit 2. Input of the training data set is
represented as a binary matrix with rows as the number of inputs and columns
being the number of digits in the input string. Population is represented as
a ternary condition string and class tuple. The conditions are represented as
a ternary matrix.

Scripts in the release
---------------------

sampleTrainScript.m provides an example of using the UCS to train and test on datasets.

ucstrain.m trains the UCS.

ucspredict.m predicts the class of an input given the trained UCS model.

match.m matches the rules against a given input.
