Population Size in Classifier Systems
Abstract
The question of optimal population size in Learning Classifier Systems is analyzed empirically in this paper. Classifier Systems are rule-based systems that use Genetic Algorithms, a model of natural selection and genetics, as their principal learning mechanism, and an economic model as their principal apportionment of credit mechanism. *CFS is a parallel Classifier System implemented on the Connection Machine. The implementation of *CFS demonstrates the validity of the long held belief that Classifier Systems and Genetic Algorithms are inherently parallel, and allows the exploration of large scale tasks that are not practical on serial computers. Published theory states that there is an optimal population size for pure Genetic Algorithms. This paper describes the difficulties in applying that theory to Classifier Systems, and presents empirical evidence that suggests there is no optimal population size for Classifier Systems.
Cite
Text
Robertson. "Population Size in Classifier Systems." International Conference on Machine Learning, 1988. doi:10.1016/B978-0-934613-64-4.50020-7Markdown
[Robertson. "Population Size in Classifier Systems." International Conference on Machine Learning, 1988.](https://mlanthology.org/icml/1988/robertson1988icml-population/) doi:10.1016/B978-0-934613-64-4.50020-7BibTeX
@inproceedings{robertson1988icml-population,
title = {{Population Size in Classifier Systems}},
author = {Robertson, George G.},
booktitle = {International Conference on Machine Learning},
year = {1988},
pages = {142-152},
doi = {10.1016/B978-0-934613-64-4.50020-7},
url = {https://mlanthology.org/icml/1988/robertson1988icml-population/}
}