Analysis of Population Evolution in Classifier Systems Using Symbolic Representations

Authors:
Pier Luca Lanzi;Stefano Rocca;Kumara Sastry;Stefania Solari
Affiliations:
Artificial Intelligence and Robotics Laboratory (AIRLab), Politecnico di Milano, Milano, Italy I-20133 and Illinois Genetic Algorithm Laboratory, University of Illinois at Urbana-Champaign, Urbana ...;Artificial Intelligence and Robotics Laboratory (AIRLab), Politecnico di Milano, Milano, Italy I-20133;Illinois Genetic Algorithm Laboratory, University of Illinois at Urbana-Champaign, Urbana, Illinois, USA;Artificial Intelligence and Robotics Laboratory (AIRLab), Politecnico di Milano, Milano, Italy I-20133
Venue:
Learning Classifier Systems
Year:
2008

Citing 8
Cited 1

Proceedings of the third international conference on Genetic algorithms

Proceedings of the third international conference on Genetic algorithms
Representational difficulties with classifier systems

Proceedings of the third international conference on Genetic algorithms
VCS: variable classifier systems

Proceedings of the third international conference on Genetic algorithms
A tale of two representations

IEA/AIE '94 Proceedings of the 7th international conference on Industrial and engineering applications of artificial intelligence and expert systems
Mining Oblique Data with XCS

IWLCS '00 Revised Papers from the Third International Workshop on Advances in Learning Classifier Systems
A comparison of bloat control methods for genetic programming

Evolutionary Computation
Generalisation of the limiting distribution of program sizes in tree-based genetic programming and analysis of its effects on bloat

Proceedings of the 9th annual conference on Genetic and evolutionary computation
A simple but theoretically-motivated method to control bloat in genetic programming

EuroGP'03 Proceedings of the 6th European conference on Genetic programming

Dynamical genetic programming in xcsf

Evolutionary Computation

Quantified Score

Hi-index	0.01

Visualization

Abstract

This paper presents an approach to analyze population evolution in classifier systems using a symbolic representation. Given a sequence of populations, representing the evolution of a solution, the method simplifies the classifiers in the populations by reducing them to their "canonical form". Then, it extracts all the subexpressions that appear in all the classifier conditions and, for each subexpression, it computes the number of occurrences in each population. Finally, it computes the trend of all the subexpressions considered. The expressions which show an increasing trend through the course of evolution are viewed as building blocks that the system has used to construct the solution.