A Robust Approach to Numeric Discovery
Abstract
IDS is an integrated discovery system that forms taxonomic hierarchies, notes qualitative relations, and finds numeric laws. This paper focuses on the system's algorithm for numeric discovery. We describe the basic method in terms of heuristic search through a space of numeric terms, provide a simple example, and show how IDS uses the algorithm to find three classes of numeric relations. We then evaluate the system's law-finding ability through experiments with artificial domains, examining the effect of system parameters, noise level, number of irrelevant terms, and law complexity on both asymptotic accuracy and learning rate. Finally, we consider the algorithm's relation to other numeric methods and outline directions for future work.
Cite
Text
Nordhausen and Langley. "A Robust Approach to Numeric Discovery." International Conference on Machine Learning, 1990. doi:10.1016/B978-1-55860-141-3.50052-3Markdown
[Nordhausen and Langley. "A Robust Approach to Numeric Discovery." International Conference on Machine Learning, 1990.](https://mlanthology.org/icml/1990/nordhausen1990icml-robust/) doi:10.1016/B978-1-55860-141-3.50052-3BibTeX
@inproceedings{nordhausen1990icml-robust,
title = {{A Robust Approach to Numeric Discovery}},
author = {Nordhausen, Bernd and Langley, Pat},
booktitle = {International Conference on Machine Learning},
year = {1990},
pages = {411-418},
doi = {10.1016/B978-1-55860-141-3.50052-3},
url = {https://mlanthology.org/icml/1990/nordhausen1990icml-robust/}
}