ML Anthology

Measuring Utility and the Design of Provably Good EBL Algorithms

Devika Subramanian, Scott B. Hunter
ICML 1992 pp. 426-425
doi:10.1016/B978-1-55860-247-2.50060-7 /icml/1992/subramanian1992icml-measuring/

Abstract

We estimate the utility of theory modifications constructed by explanation-based learning over given problem distributions. This paper describes an algorithm for efficiently estimating the cost of solving a distribution Q of queries in redundant, as well as recursive theories. The cost estimates produced by the algorithm are within user-specified error (ε)and confidence (δ) bounds. We use the estimation theory to determine a key property called separability, that ensures that redundant macros speed up the execution of query distributions. We derive and empirically validate two utility theorems on EBL.

ICML Semantic Scholar

Cite

Text

Subramanian and Hunter. "Measuring Utility and the Design of Provably Good EBL Algorithms." International Conference on Machine Learning, 1992. doi:10.1016/B978-1-55860-247-2.50060-7

Markdown

[Subramanian and Hunter. "Measuring Utility and the Design of Provably Good EBL Algorithms." International Conference on Machine Learning, 1992.](https://mlanthology.org/icml/1992/subramanian1992icml-measuring/) doi:10.1016/B978-1-55860-247-2.50060-7

BibTeX

@inproceedings{subramanian1992icml-measuring,
  title     = {{Measuring Utility and the Design of Provably Good EBL Algorithms}},
  author    = {Subramanian, Devika and Hunter, Scott B.},
  booktitle = {International Conference on Machine Learning},
  year      = {1992},
  pages     = {426-425},
  doi       = {10.1016/B978-1-55860-247-2.50060-7},
  url       = {https://mlanthology.org/icml/1992/subramanian1992icml-measuring/}
}