Optimal and Adaptive Algorithms for Online Boosting

Alina Beygelzimer, Satyen Kale, Haipeng Luo
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:2323-2331, 2015.

Abstract

We study online boosting, the task of converting any weak online learner into a strong online learner. Based on a novel and natural definition of weak online learnability, we develop two online boosting algorithms. The first algorithm is an online version of boost-by-majority. By proving a matching lower bound, we show that this algorithm is essentially optimal in terms of the number of weak learners and the sample complexity needed to achieve a specified accuracy. The second algorithm is adaptive and parameter-free, albeit not optimal.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-beygelzimer15, title = {Optimal and Adaptive Algorithms for Online Boosting}, author = {Beygelzimer, Alina and Kale, Satyen and Luo, Haipeng}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {2323--2331}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/beygelzimer15.pdf}, url = {https://proceedings.mlr.press/v37/beygelzimer15.html}, abstract = {We study online boosting, the task of converting any weak online learner into a strong online learner. Based on a novel and natural definition of weak online learnability, we develop two online boosting algorithms. The first algorithm is an online version of boost-by-majority. By proving a matching lower bound, we show that this algorithm is essentially optimal in terms of the number of weak learners and the sample complexity needed to achieve a specified accuracy. The second algorithm is adaptive and parameter-free, albeit not optimal.} }
Endnote
%0 Conference Paper %T Optimal and Adaptive Algorithms for Online Boosting %A Alina Beygelzimer %A Satyen Kale %A Haipeng Luo %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-beygelzimer15 %I PMLR %P 2323--2331 %U https://proceedings.mlr.press/v37/beygelzimer15.html %V 37 %X We study online boosting, the task of converting any weak online learner into a strong online learner. Based on a novel and natural definition of weak online learnability, we develop two online boosting algorithms. The first algorithm is an online version of boost-by-majority. By proving a matching lower bound, we show that this algorithm is essentially optimal in terms of the number of weak learners and the sample complexity needed to achieve a specified accuracy. The second algorithm is adaptive and parameter-free, albeit not optimal.
RIS
TY - CPAPER TI - Optimal and Adaptive Algorithms for Online Boosting AU - Alina Beygelzimer AU - Satyen Kale AU - Haipeng Luo BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-beygelzimer15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 2323 EP - 2331 L1 - http://proceedings.mlr.press/v37/beygelzimer15.pdf UR - https://proceedings.mlr.press/v37/beygelzimer15.html AB - We study online boosting, the task of converting any weak online learner into a strong online learner. Based on a novel and natural definition of weak online learnability, we develop two online boosting algorithms. The first algorithm is an online version of boost-by-majority. By proving a matching lower bound, we show that this algorithm is essentially optimal in terms of the number of weak learners and the sample complexity needed to achieve a specified accuracy. The second algorithm is adaptive and parameter-free, albeit not optimal. ER -
APA
Beygelzimer, A., Kale, S. & Luo, H.. (2015). Optimal and Adaptive Algorithms for Online Boosting. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:2323-2331 Available from https://proceedings.mlr.press/v37/beygelzimer15.html.

Related Material