Model Induction: A New Source of CSP Model Redundancy
Abstract
Based on the notions of viewpoints, models, and channeling constraints, the paper introduces model induction, a systematic transformation of constraints in an existing model to constraints in another viewpoint. Meant to be a general CSP model operator, model induction is useful in generating redundant models, which can be further induced or combined with the original model or other mutually redundant models. We propose three ways of combining redundant models using model induction, model channeling, and model intersection. Experimental results on the Langford's problem confirm that our proposed combined models exhibit improvements in efficiency and robustness over the original single models.
Cite
Text
Law and Lee. "Model Induction: A New Source of CSP Model Redundancy." AAAI Conference on Artificial Intelligence, 2002. doi:10.5555/777092.777104Markdown
[Law and Lee. "Model Induction: A New Source of CSP Model Redundancy." AAAI Conference on Artificial Intelligence, 2002.](https://mlanthology.org/aaai/2002/law2002aaai-model/) doi:10.5555/777092.777104BibTeX
@inproceedings{law2002aaai-model,
title = {{Model Induction: A New Source of CSP Model Redundancy}},
author = {Law, Yat Chiu and Lee, Jimmy Ho-Man},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2002},
pages = {54-61},
doi = {10.5555/777092.777104},
url = {https://mlanthology.org/aaai/2002/law2002aaai-model/}
}