Cost-Sensitive Tree of Classifiers

Zhixiang Xu, Matt Kusner, Kilian Weinberger, Minmin Chen
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(1):133-141, 2013.

Abstract

Recently, machine learning algorithms have successfully entered large-scale real-world industrial applications (e.g. search engines and email spam filters). Here, the CPU cost during test-time must be budgeted and accounted for. In this paper, we address the challenge of balancing test-time cost and the classifier accuracy in a principled fashion. The test-time cost of a classifier is often dominated by the computation required for feature extraction-which can vary drastically across features. We incorporate this extraction time by constructing a tree of classifiers, through which test inputs traverse along individual paths. Each path extracts different features and is optimized for a specific sub-partition of the input space. By only computing features for inputs that benefit from them the most, our cost-sensitive tree of classifiers can match the high accuracies of the current state-of-the-art at a small fraction of the computational cost.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-xu13, title = {Cost-Sensitive Tree of Classifiers}, author = {Xu, Zhixiang and Kusner, Matt and Weinberger, Kilian and Chen, Minmin}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {133--141}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/xu13.pdf}, url = {https://proceedings.mlr.press/v28/xu13.html}, abstract = {Recently, machine learning algorithms have successfully entered large-scale real-world industrial applications (e.g. search engines and email spam filters). Here, the CPU cost during test-time must be budgeted and accounted for. In this paper, we address the challenge of balancing test-time cost and the classifier accuracy in a principled fashion. The test-time cost of a classifier is often dominated by the computation required for feature extraction-which can vary drastically across features. We incorporate this extraction time by constructing a tree of classifiers, through which test inputs traverse along individual paths. Each path extracts different features and is optimized for a specific sub-partition of the input space. By only computing features for inputs that benefit from them the most, our cost-sensitive tree of classifiers can match the high accuracies of the current state-of-the-art at a small fraction of the computational cost.} }
Endnote
%0 Conference Paper %T Cost-Sensitive Tree of Classifiers %A Zhixiang Xu %A Matt Kusner %A Kilian Weinberger %A Minmin Chen %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-xu13 %I PMLR %P 133--141 %U https://proceedings.mlr.press/v28/xu13.html %V 28 %N 1 %X Recently, machine learning algorithms have successfully entered large-scale real-world industrial applications (e.g. search engines and email spam filters). Here, the CPU cost during test-time must be budgeted and accounted for. In this paper, we address the challenge of balancing test-time cost and the classifier accuracy in a principled fashion. The test-time cost of a classifier is often dominated by the computation required for feature extraction-which can vary drastically across features. We incorporate this extraction time by constructing a tree of classifiers, through which test inputs traverse along individual paths. Each path extracts different features and is optimized for a specific sub-partition of the input space. By only computing features for inputs that benefit from them the most, our cost-sensitive tree of classifiers can match the high accuracies of the current state-of-the-art at a small fraction of the computational cost.
RIS
TY - CPAPER TI - Cost-Sensitive Tree of Classifiers AU - Zhixiang Xu AU - Matt Kusner AU - Kilian Weinberger AU - Minmin Chen BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/02/13 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-xu13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 1 SP - 133 EP - 141 L1 - http://proceedings.mlr.press/v28/xu13.pdf UR - https://proceedings.mlr.press/v28/xu13.html AB - Recently, machine learning algorithms have successfully entered large-scale real-world industrial applications (e.g. search engines and email spam filters). Here, the CPU cost during test-time must be budgeted and accounted for. In this paper, we address the challenge of balancing test-time cost and the classifier accuracy in a principled fashion. The test-time cost of a classifier is often dominated by the computation required for feature extraction-which can vary drastically across features. We incorporate this extraction time by constructing a tree of classifiers, through which test inputs traverse along individual paths. Each path extracts different features and is optimized for a specific sub-partition of the input space. By only computing features for inputs that benefit from them the most, our cost-sensitive tree of classifiers can match the high accuracies of the current state-of-the-art at a small fraction of the computational cost. ER -
APA
Xu, Z., Kusner, M., Weinberger, K. & Chen, M.. (2013). Cost-Sensitive Tree of Classifiers. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(1):133-141 Available from https://proceedings.mlr.press/v28/xu13.html.

Related Material