Generalization in Wilson's classifier system

  • Pier Luca Lanzi
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1498)


We analyze generalization with the XCS classifier system when the system is applied to animat problems in grid-worlds. Our aim is to give a unified view of generalization with XCS, in order to explain some of the phenomena reported in the literature. Initially, we apply XCS to two environments. Our results show that there are situations in which the generalization mechanism of XCS may prevent the system from converging to optimum. Accordingly, we study XCS's generalization mechanism analyzing the conditions under which the system may fail to evolve an optimal solution. We draw a hypothesis in order to explain the results reported so far. Our hypothesis suggests that XCS fails to learn an optimal solution when, due to the environment structure and to the exploration strategy employed, the system does not explore all the areas of the environment frequently. We thus introduce a meta exploration strategy that is used as theoretical tool to validate our hypothesis experimentally.


Optimal Policy Classifier System Exploration Strategy Theoretical Tool Environmental Niche 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Dave Cliff and Susi Ross. Adding memory to ZCS. Adaptive Behaviour, 3(2):101–150, 1994.Google Scholar
  2. 2.
    Pier Luca Lanzi. A Model of the Environment to Avoid Local Learning (An Analysis of the Generalization Mechanism of XCS). Technical Report 97.46, Dipartimento di Elettronica e Informazione — Politecnico di Milano, 1997. Available at Scholar
  3. 3.
    Pier Luca Lanzi. Solving Problems in Partially Observable Environments with Classifier Systems (Experiments on Adding Memory to XCS). Technical Report 97.45, Dipartimento di Elettronica e Informazione — Politecnico di Milano, 1997. Available at Scholar
  4. 4.
    Pier Luca Lanzi. A Study on the Generalization Capabilities of XCS. In Proceedings of the Seventh International Conference on Genetic Algorithms. Morgan Kaufmann, 1997.Google Scholar
  5. 5.
    Stewart W. Wilson. ZCS: a zeroth level classifier system. Evolutionary Computation, 1(2):1–18, 1994.Google Scholar
  6. 6.
    Stewart W. Wilson. Classifier fitness based on accuracy. Evolutionary Computation, 3(2):149–175, 1995.Google Scholar
  7. 7.
    Stewart W. Wilson. Personal communication. 1997.Google Scholar
  8. 8.
    Stewart W. Wilson. Generalization in the XCS classifier system. In MIT Press, editor, Proceedings of the Third Annual Genetic Programming Conference (GP-98), 1998.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1998

Authors and Affiliations

  • Pier Luca Lanzi
    • 1
  1. 1.Artificial Intelligence & Robotics Project Dipartimento di Elettronica e InformazionePolitecnico di MilanoMilanoItaly

Personalised recommendations