Learning Classifier Systems

The decision-making engine in the CS is the set of classifier condition-action rules therefore, the key to a successful application is a well-constructed set of rules. If the control problem is straightforward, the necessary classifiers could, in principle, be created by hand, but there is rarely much point in doing this. A single classifier is equivalent to a production rule, the same structures that form the basis of most expert systems if a set of classifiers that could adequately control the environment could be created by hand, it would probably be as easy to create an equivalent expert system (ES). As an ES is able to explain its actions but a CS is not, in these circumstances, an ES would be preferable. 

The CS comes into its own when the rules that are needed to control the environment are only partly known, or are completely unknown, so that a comprehensive set of ES rules cannot be constructed by hand. If we cannot create the ES by hand, it must also be impossible to create the CS by hand some other method of creating the classifiers must be found. This is the realm of the learning classifier system (LCS) in which all classifier systems of value lie. 

To make sense out of this chaos, a method is needed to pick out the rules that will be useful in controlling the system from among the large number that are worthless. The process of identifying productive classifiers relies on a mechanism that provides rewards to those rules that are helpful with a penalty for those that are not. The better rules then gradually emerge from the background noise. 

---

This software model is a learning classifier system. Because classifier systems learn, they can be applied to the control of a dynamic system, such as a reactor or an instrument, which must process various types of samples under unpredictable conditions, even when the rules required for successful control are unknown. 

A helpful starting point for further investigation is Learning Classifier Systems From Foundations to Applications.1 The literature in classifier systems is far thinner than that in genetic algorithms, artificial neural networks, and other methods discussed in this book. A productive way to uncover more... 

Lanzi, P.L., Stolzmann, W., and Wilson, S.W., (Eds.) Learning classifier systems From foundations to applications, Lecture Notes in Artificial Intelligence 1813, Springer, Berlin, 2000. 

Discusses evolutionary algorithms, cellular automata, expert systems, fuzzy logic, learning classifier systems, and evolvable developmental systems... 

These results form the basis for the development of a learning classifier system (22,23) for combustion control in multiple burner installations. 

Learning classifier systems that use agents representing set of rules as a solution to machine learning problem. 

---