Nonparametric Classification with Polynomial MPMC Cascades
Abstract
A new class of nonparametric algorithms for high-dimensional binary classification is proposed using cascades of low dimensional polynomial structures. Construction of polynomial cascadesis based on Minimax Probability Machine Classification (MPMC), which results in direct estimates of classificationaccuracy, and provides a simple stopping criteria that does not requireexpensive cross-validation measures. This Polynomial MPMC Cascade (PMC) algorithm is constructed in linear timewith respect to the input space dimensionality, and linear time in the numberof examples, making it a potentially attractive alternative to algorithms likesupport vector machines and standard MPMC. Experimental evidence is given showing that, compared to state-of-the-artclassifiers, PMCs are competitive; inherently fast to compute; not prone tooverfitting; and generally yield accurate estimates of the maximum error rateon unseen data.
Cite
Text
Bohté et al. "Nonparametric Classification with Polynomial MPMC Cascades." International Conference on Machine Learning, 2004. doi:10.1145/1015330.1015416Markdown
[Bohté et al. "Nonparametric Classification with Polynomial MPMC Cascades." International Conference on Machine Learning, 2004.](https://mlanthology.org/icml/2004/bohte2004icml-nonparametric/) doi:10.1145/1015330.1015416BibTeX
@inproceedings{bohte2004icml-nonparametric,
title = {{Nonparametric Classification with Polynomial MPMC Cascades}},
author = {Bohté, Sander M. and Breitenbach, Markus and Grudic, Gregory Z.},
booktitle = {International Conference on Machine Learning},
year = {2004},
doi = {10.1145/1015330.1015416},
url = {https://mlanthology.org/icml/2004/bohte2004icml-nonparametric/}
}