Tuning Rule-Based Systems to Their Environments
Abstract
Lamp presents a general-purpose, knowledge-poor learning strategy for incrementally improving the performance of rule-based systems operating in reactive environments. The LAMP critic observes the performance of an MEA planner vis-a-vis a set of computed relations that define the interface between the real world and the reasoning engine's model of that world. Planning successes and failures are traced back through action definitions to the underlying, primitive relations. A domain-independent success criterion describes the performance of the system and suggests modifications to the relation definitions in order to tune the system to the particular environment and task.
Cite
Text
Tallis. "Tuning Rule-Based Systems to Their Environments." International Conference on Machine Learning, 1988. doi:10.1016/B978-0-934613-64-4.50005-0Markdown
[Tallis. "Tuning Rule-Based Systems to Their Environments." International Conference on Machine Learning, 1988.](https://mlanthology.org/icml/1988/tallis1988icml-tuning/) doi:10.1016/B978-0-934613-64-4.50005-0BibTeX
@inproceedings{tallis1988icml-tuning,
title = {{Tuning Rule-Based Systems to Their Environments}},
author = {Tallis, Hans},
booktitle = {International Conference on Machine Learning},
year = {1988},
pages = {8-14},
doi = {10.1016/B978-0-934613-64-4.50005-0},
url = {https://mlanthology.org/icml/1988/tallis1988icml-tuning/}
}