Adversarial Cost-Sensitive Classification

Kaiser Asif U of Illinois at Chicago, Wei Xing U of Illinois at Chicago, Sima Behpour U of Illinois at Chicago, Brian Ziebart
Proceedings of the 31st Conference on Uncertainty in Artificial Intelligence, PMLR R13:59-68, 2015.

Abstract

In many classification settings, mistakes incur different application-dependent penalties based on the predicted and the actual class label. Cost-sensitive classifiers that attempt to minimize these application-based penalties are needed. We propose a robust minimax approach for producing classifiers that directly minimize the cost of mistakes as a convex optimization problem. This is in contrast to previous methods that minimize the empirical risk using a convex surrogate for the cost of mistakes, since minimizing the empirical risk of the actual cost-sensitive loss is generally intractable. By treating properties of the training data as being uncertain, our approach avoids these computational difficulties. We develop theory and algorithms for our approach and demonstrate its benefits on cost-sensitive classification tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR13-chicago15a, title = {Adversarial Cost-Sensitive Classification}, author = {Chicago, Kaiser Asif U of Illinois at and Chicago, Wei Xing U of Illinois at and Chicago, Sima Behpour U of Illinois at and Ziebart, Brian}, booktitle = {Proceedings of the 31st Conference on Uncertainty in Artificial Intelligence}, pages = {59--68}, year = {2015}, editor = {Meila, Marina and Heskes, Tom}, volume = {R13}, series = {Proceedings of Machine Learning Research}, month = {12--16 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r13/main/assets/chicago15a/chicago15a.pdf}, url = {https://proceedings.mlr.press/r13/chicago15a.html}, abstract = {In many classification settings, mistakes incur different application-dependent penalties based on the predicted and the actual class label. Cost-sensitive classifiers that attempt to minimize these application-based penalties are needed. We propose a robust minimax approach for producing classifiers that directly minimize the cost of mistakes as a convex optimization problem. This is in contrast to previous methods that minimize the empirical risk using a convex surrogate for the cost of mistakes, since minimizing the empirical risk of the actual cost-sensitive loss is generally intractable. By treating properties of the training data as being uncertain, our approach avoids these computational difficulties. We develop theory and algorithms for our approach and demonstrate its benefits on cost-sensitive classification tasks.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Adversarial Cost-Sensitive Classification %A Kaiser Asif U of Illinois at Chicago %A Wei Xing U of Illinois at Chicago %A Sima Behpour U of Illinois at Chicago %A Brian Ziebart %B Proceedings of the 31st Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2015 %E Marina Meila %E Tom Heskes %F pmlr-vR13-chicago15a %I PMLR %P 59--68 %U https://proceedings.mlr.press/r13/chicago15a.html %V R13 %X In many classification settings, mistakes incur different application-dependent penalties based on the predicted and the actual class label. Cost-sensitive classifiers that attempt to minimize these application-based penalties are needed. We propose a robust minimax approach for producing classifiers that directly minimize the cost of mistakes as a convex optimization problem. This is in contrast to previous methods that minimize the empirical risk using a convex surrogate for the cost of mistakes, since minimizing the empirical risk of the actual cost-sensitive loss is generally intractable. By treating properties of the training data as being uncertain, our approach avoids these computational difficulties. We develop theory and algorithms for our approach and demonstrate its benefits on cost-sensitive classification tasks. %Z Reissued by PMLR on 04 October 2026.
APA
Chicago, K.A.U.o.I.a., Chicago, W.X.U.o.I.a., Chicago, S.B.U.o.I.a. & Ziebart, B.. (2015). Adversarial Cost-Sensitive Classification. Proceedings of the 31st Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R13:59-68 Available from https://proceedings.mlr.press/r13/chicago15a.html. Reissued by PMLR on 04 October 2026.

Related Material