Toward Optimal Classifier System Performance in Non-Markov Environments

Article Properties
  • Language
    English
  • Publication Date
    2000/12/01
  • Indian UGC (Journal)
  • Refrences
    9
  • Citations
    22
  • Pier Luca Lanzi Artificial Intelligence and, Robotics Laboratory, Dip. di Elettronica e Informazione, Politecnico di Milano, Piazza Leonardo da Vinci n. 32, I-20133 Milano, Italy
  • Stewart W. Wilson Department of General Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801-2996, USA
Abstract
Cite
Lanzi, Pier Luca, and Stewart W. Wilson. “Toward Optimal Classifier System Performance in Non-Markov Environments”. Evolutionary Computation, vol. 8, no. 4, 2000, pp. 393-18, https://doi.org/10.1162/106365600568239.
Lanzi, P. L., & Wilson, S. W. (2000). Toward Optimal Classifier System Performance in Non-Markov Environments. Evolutionary Computation, 8(4), 393-418. https://doi.org/10.1162/106365600568239
Lanzi PL, Wilson SW. Toward Optimal Classifier System Performance in Non-Markov Environments. Evolutionary Computation. 2000;8(4):393-418.
Journal Categories
Science
Mathematics
Instruments and machines
Electronic computers
Computer science
Science
Mathematics
Instruments and machines
Electronic computers
Computer science
Computer software
Technology
Electrical engineering
Electronics
Nuclear engineering
Electronics
Computer engineering
Computer hardware
Description

Can a classifier system achieve near-optimal performance in complex environments? This research explores the incorporation of Wilson's bit-register memory scheme into the XCS classifier system within non-Markov environments. The system is designed to handle situations where the current state does not fully predict future states. Two key extensions were critical for achieving near-optimal performance in difficult environments. The first was an exploration strategy where external actions were probabilistic but internal “actions” were deterministic. The second involved using a register with more bit-positions than necessary to resolve environmental aliasing. The study discusses the origins and effects of these two extensions, demonstrating their impact on the classifier's ability to learn and adapt. These findings contribute to the advancement of classifier systems and their ability to solve complex problems in real-world scenarios. The insights gained could be valuable in designing more robust and adaptable AI agents for applications such as robotics, control systems, and data mining, enhancing performance in environments with incomplete information.

Published in Evolutionary Computation, this paper fits into the journal's focus on adaptive systems and machine learning. The use of a classifier system, a type of evolutionary algorithm, is consistent with the journal's scope. The paper contributes to the understanding of how evolutionary computation methods can be applied to solve problems in complex and uncertain environments, which is of interest to the journal's readership.

Refrences
Citations
Citations Analysis
The first research to cite this article was titled ZCS Redux and was published in 2002. The most recent citation comes from a 2022 study titled ZCS Redux . This article reached its peak citation in 2015 , with 4 citations.It has been cited in 14 different journals. Among related journals, the Evolutionary Intelligence cited this research the most, with 5 citations. The chart below illustrates the annual citation trends for this article.
Citations used this article by year