Search Control, Utility, and Concept Induction
Abstract
Our research adapts incremental conceptual clustering (or concept formation) to the task of learning to guide search. We build on earlier research that uses concept induction techniques to learn search control, but our approach differs by virtue of its reliance on probabilistic, hierarchical classification schemes that increase certain aspects of search efficiency. The system also includes inductive strategies of ‘noise tolerance’ that mitigate problems of control knowledge ‘utility’. A general lesson is that recently identified search ‘utility’ problems are synonymous with inductive problems of ‘noise’; solutions to the problems of the latter type can be usefully adapted to the former.
Cite
Text
Carlson et al. "Search Control, Utility, and Concept Induction." International Conference on Machine Learning, 1990. doi:10.1016/B978-1-55860-141-3.50014-6Markdown
[Carlson et al. "Search Control, Utility, and Concept Induction." International Conference on Machine Learning, 1990.](https://mlanthology.org/icml/1990/carlson1990icml-search/) doi:10.1016/B978-1-55860-141-3.50014-6BibTeX
@inproceedings{carlson1990icml-search,
title = {{Search Control, Utility, and Concept Induction}},
author = {Carlson, Brian M. and Weinberg, Jerry B. and Fisher, Douglas H.},
booktitle = {International Conference on Machine Learning},
year = {1990},
pages = {85-92},
doi = {10.1016/B978-1-55860-141-3.50014-6},
url = {https://mlanthology.org/icml/1990/carlson1990icml-search/}
}