Automatic Feature Generation for Problem Solving Systems

Abstract

Existing methods for constructive induction generally isolate feature generation from problem solving. This paper describes a theory of feature generation that creates features using both a domain theory and feedback performance. An evaluation function based on these features is learned simultaneously and is used to guide a problem solver. The Zenith system, an implementation of this theory, has been applied to two domains. In each domain, Zenith generated useful features, given only a domain theory and the ability to solve problems in the domain.

Cite

Text

Fawcett and Utgoff. "Automatic Feature Generation for Problem Solving Systems." International Conference on Machine Learning, 1992. doi:10.1016/B978-1-55860-247-2.50024-3

Markdown

[Fawcett and Utgoff. "Automatic Feature Generation for Problem Solving Systems." International Conference on Machine Learning, 1992.](https://mlanthology.org/icml/1992/fawcett1992icml-automatic/) doi:10.1016/B978-1-55860-247-2.50024-3

BibTeX

@inproceedings{fawcett1992icml-automatic,
  title     = {{Automatic Feature Generation for Problem Solving Systems}},
  author    = {Fawcett, Tom and Utgoff, Paul E.},
  booktitle = {International Conference on Machine Learning},
  year      = {1992},
  pages     = {144-153},
  doi       = {10.1016/B978-1-55860-247-2.50024-3},
  url       = {https://mlanthology.org/icml/1992/fawcett1992icml-automatic/}
}