Research on learning classifier systems based on genetic algorithm

The problems of the learning classifier system are discussed based on genetic algorithm. The system framework and the key techniques are analyzed, such as messages and classifiers match, credit assignment with bucket-brigade algorithm, rule discovery with genetic algorithm. The convergence for the strength update of classifiers is proved with bucket-brigade algorithm. The simulation results on the system performance and the convergence of the strength update are obtained under the conditions of the 6-multiplexer Boolean function learning.