Fringe-like Feature Construction: A Comparative Study and a Unifying Scheme
Abstract
A class of concept learning algorithms CL augments standard similarity-based techniques by performing feature construction based on the SBL output. Three instances of this class—Pagallo and Haussler's FRINGE, Pagallo's extension Symmetric FRINGE (SymFringe) and a refinement DCFringe—use patterns at the fringe of a decision tree to guide their construction, but DCFringe uses limited construction of conjunction and disjunction. Experiments with small DNF and CNF concepts show that DCFringe outperforms both FRINGE and SymFringe, in terms of accuracy, conciseness, and efficiency. We discuss limitations of current methods, which leads to a feature construction approach based on a wider variety of patterns restricted by statistical measures and optional knowledge.
Cite
Text
Yang et al. "Fringe-like Feature Construction: A Comparative Study and a Unifying Scheme." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50048-9Markdown
[Yang et al. "Fringe-like Feature Construction: A Comparative Study and a Unifying Scheme." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/yang1991icml-fringe/) doi:10.1016/B978-1-55860-200-7.50048-9BibTeX
@inproceedings{yang1991icml-fringe,
title = {{Fringe-like Feature Construction: A Comparative Study and a Unifying Scheme}},
author = {Yang, Der-Shung and Rendell, Larry A. and Blix, Gunnar},
booktitle = {International Conference on Machine Learning},
year = {1991},
pages = {223-227},
doi = {10.1016/B978-1-55860-200-7.50048-9},
url = {https://mlanthology.org/icml/1991/yang1991icml-fringe/}
}