Interpretable Two-level Boolean Rule Learning for Classification
Abstract
As a contribution to interpretable machine learning research, we develop a novel optimization framework for learning accurate and sparse two-level Boolean rules. We consider rules in both conjunctive normal form (AND-of-ORs) and disjunctive normal form (OR-of-ANDs). A principled objective function is proposed to trade classification accuracy and interpretability, where we use Hamming loss to characterize accuracy and sparsity to characterize interpretability. We propose efficient procedures to optimize these objectives based on linear programming (LP) relaxation, block coordinate descent, and alternating minimization. Experiments show that our new algorithms provide very good tradeoffs between accuracy and interpretability.
Cite
@article{arxiv.1606.05798,
title = {Interpretable Two-level Boolean Rule Learning for Classification},
author = {Guolong Su and Dennis Wei and Kush R. Varshney and Dmitry M. Malioutov},
journal= {arXiv preprint arXiv:1606.05798},
year = {2016}
}
Comments
presented at 2016 ICML Workshop on Human Interpretability in Machine Learning (WHI 2016), New York, NY