Learning Variable Descriptors for Applying Heuristics Across CSP Problems
Abstract
Many naturally occurring problems can be usefully represented as constraint satisfaction problems for which a variety of general purpose algorithms are available. However, to date the ability to improve problem solving performance with experience has been largely limited to learning constraints to prune the search space or other value-ordering heuristics whose applicability is restricted to completely identical constraint networks. In this research we show how value-ordering heuristics can be formed using a language that allows for their application across very different problem instances. This language is formed incrementally on the basis of the variables seen in the course of solving different problems.
Cite
Text
Day. "Learning Variable Descriptors for Applying Heuristics Across CSP Problems." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50029-5Markdown
[Day. "Learning Variable Descriptors for Applying Heuristics Across CSP Problems." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/day1991icml-learning/) doi:10.1016/B978-1-55860-200-7.50029-5BibTeX
@inproceedings{day1991icml-learning,
title = {{Learning Variable Descriptors for Applying Heuristics Across CSP Problems}},
author = {Day, David S.},
booktitle = {International Conference on Machine Learning},
year = {1991},
pages = {127-131},
doi = {10.1016/B978-1-55860-200-7.50029-5},
url = {https://mlanthology.org/icml/1991/day1991icml-learning/}
}