Ideal Theory Refinement Under Object Identity
Abstract
We present a framework for theory refinement operators fulfilling some desirable properties in order to ensure the efficiency and effectiveness of the process. A refinement operator satisfying these requirements is defined ideal. In other frameworks, the search for refinements takes place in spaces ordered by the logical implication or the theta-subsumption relationships. Results have demonstrated the impossibility of defining ideal operators in these search spaces. We show that, by assuming the object identity bias over a clausal representation, we are able to define them in the resulting search space.
Cite
Text
Esposito et al. "Ideal Theory Refinement Under Object Identity." International Conference on Machine Learning, 2000.Markdown
[Esposito et al. "Ideal Theory Refinement Under Object Identity." International Conference on Machine Learning, 2000.](https://mlanthology.org/icml/2000/esposito2000icml-ideal/)BibTeX
@inproceedings{esposito2000icml-ideal,
title = {{Ideal Theory Refinement Under Object Identity}},
author = {Esposito, Floriana and Fanizzi, Nicola and Ferilli, Stefano and Semeraro, Giovanni},
booktitle = {International Conference on Machine Learning},
year = {2000},
pages = {263-270},
url = {https://mlanthology.org/icml/2000/esposito2000icml-ideal/}
}