An Empirical Analysis of EBL Approaches for Learning Plan Schemata
Abstract
One approach to the intractability of planning is to build a schema-based planner. When the planner is given a goal to achieve, it simply looks for a solution in its schema library. Once a schema is selected, several actions can be incorporated into a plan without the need for intervening search. This paper investigates the use of explanation-based learning (EBL) to acquire schemata for such a planner. An empirical comparison of two approaches for acquiring plan schemata is reported. The results demonstrate the advantages of learning by observing the intelligent behavior of external agents and illustrate the weaknesses of having a system learn from its own problem solving. The experiments also compare two styles of EBL algorithms and demonstrate the efficacy of EBL in general. Suggestions are made about the appropriate design of practical systems that learn plan schemata.
Cite
Text
Shavlik. "An Empirical Analysis of EBL Approaches for Learning Plan Schemata." International Conference on Machine Learning, 1989. doi:10.1016/B978-1-55860-036-2.50052-7Markdown
[Shavlik. "An Empirical Analysis of EBL Approaches for Learning Plan Schemata." International Conference on Machine Learning, 1989.](https://mlanthology.org/icml/1989/shavlik1989icml-empirical/) doi:10.1016/B978-1-55860-036-2.50052-7BibTeX
@inproceedings{shavlik1989icml-empirical,
title = {{An Empirical Analysis of EBL Approaches for Learning Plan Schemata}},
author = {Shavlik, Jude W.},
booktitle = {International Conference on Machine Learning},
year = {1989},
pages = {183-187},
doi = {10.1016/B978-1-55860-036-2.50052-7},
url = {https://mlanthology.org/icml/1989/shavlik1989icml-empirical/}
}