ML Anthology

AGIL: Solving the Exploration Versus Exploration Dilemma in a Single Classifier System Applied to Simulated Robotics

Gilles Venturini
ICML 1992 pp. 458-463
doi:10.1016/B978-1-55860-247-2.50064-4 /icml/1992/venturini1992icml-agil/

Abstract

This paper describes an adaptive genetic learning system called AGIL that solves control problems by learning conditions-actions rules (classifiers). For this purpose, the exploration versus exploitation dilemma (choosing between using learned knowledge or learning new knowledge) is solved less sensitively to rule strength and in a more explicit and controllable way than standard genetic techniques used in simple classifier systems. AGIL can be compared favorably with other learning systems on the multiplexor function learning task. It solves a simulated task of an autonomous moving robot that must reach a target in an unknown environment with obstacles.

ICML Semantic Scholar

Cite

Text

Venturini. "AGIL: Solving the Exploration Versus Exploration Dilemma in a Single Classifier System Applied to Simulated Robotics." International Conference on Machine Learning, 1992. doi:10.1016/B978-1-55860-247-2.50064-4

Markdown

[Venturini. "AGIL: Solving the Exploration Versus Exploration Dilemma in a Single Classifier System Applied to Simulated Robotics." International Conference on Machine Learning, 1992.](https://mlanthology.org/icml/1992/venturini1992icml-agil/) doi:10.1016/B978-1-55860-247-2.50064-4

BibTeX

@inproceedings{venturini1992icml-agil,
  title     = {{AGIL: Solving the Exploration Versus Exploration Dilemma in a Single Classifier System Applied to Simulated Robotics}},
  author    = {Venturini, Gilles},
  booktitle = {International Conference on Machine Learning},
  year      = {1992},
  pages     = {458-463},
  doi       = {10.1016/B978-1-55860-247-2.50064-4},
  url       = {https://mlanthology.org/icml/1992/venturini1992icml-agil/}
}