Optimally Combining Classifiers Using Unlabeled Data

Akshay Balsubramani, Yoav Freund
Proceedings of The 28th Conference on Learning Theory, PMLR 40:211-225, 2015.

Abstract

We develop a worst-case analysis of aggregation of classifier ensembles for binary classification. The task of predicting to minimize error is formulated as a game played over a given set of unlabeled data (a transductive setting), where prior label information is encoded as constraints on the game. The minimax solution of this game identifies cases where a weighted combination of the classifiers can perform significantly better than any single classifier.

Cite this Paper

BibTeX
@InProceedings{pmlr-v40-Balsubramani15, title = {Optimally Combining Classifiers Using Unlabeled Data}, author = {Balsubramani, Akshay and Freund, Yoav}, booktitle = {Proceedings of The 28th Conference on Learning Theory}, pages = {211--225}, year = {2015}, editor = {Grünwald, Peter and Hazan, Elad and Kale, Satyen}, volume = {40}, series = {Proceedings of Machine Learning Research}, address = {Paris, France}, month = {03--06 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v40/Balsubramani15.pdf}, url = {https://proceedings.mlr.press/v40/Balsubramani15.html}, abstract = {We develop a worst-case analysis of aggregation of classifier ensembles for binary classification. The task of predicting to minimize error is formulated as a game played over a given set of unlabeled data (a transductive setting), where prior label information is encoded as constraints on the game. The minimax solution of this game identifies cases where a weighted combination of the classifiers can perform significantly better than any single classifier.} }
Endnote
%0 Conference Paper %T Optimally Combining Classifiers Using Unlabeled Data %A Akshay Balsubramani %A Yoav Freund %B Proceedings of The 28th Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2015 %E Peter Grünwald %E Elad Hazan %E Satyen Kale %F pmlr-v40-Balsubramani15 %I PMLR %P 211--225 %U https://proceedings.mlr.press/v40/Balsubramani15.html %V 40 %X We develop a worst-case analysis of aggregation of classifier ensembles for binary classification. The task of predicting to minimize error is formulated as a game played over a given set of unlabeled data (a transductive setting), where prior label information is encoded as constraints on the game. The minimax solution of this game identifies cases where a weighted combination of the classifiers can perform significantly better than any single classifier.
RIS
TY - CPAPER TI - Optimally Combining Classifiers Using Unlabeled Data AU - Akshay Balsubramani AU - Yoav Freund BT - Proceedings of The 28th Conference on Learning Theory DA - 2015/06/26 ED - Peter Grünwald ED - Elad Hazan ED - Satyen Kale ID - pmlr-v40-Balsubramani15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 40 SP - 211 EP - 225 L1 - http://proceedings.mlr.press/v40/Balsubramani15.pdf UR - https://proceedings.mlr.press/v40/Balsubramani15.html AB - We develop a worst-case analysis of aggregation of classifier ensembles for binary classification. The task of predicting to minimize error is formulated as a game played over a given set of unlabeled data (a transductive setting), where prior label information is encoded as constraints on the game. The minimax solution of this game identifies cases where a weighted combination of the classifiers can perform significantly better than any single classifier. ER -
APA
Balsubramani, A. & Freund, Y.. (2015). Optimally Combining Classifiers Using Unlabeled Data. Proceedings of The 28th Conference on Learning Theory, in Proceedings of Machine Learning Research 40:211-225 Available from https://proceedings.mlr.press/v40/Balsubramani15.html.

Related Material