Learning from Corrupted Binary Labels via Class-Probability Estimation

Aditya Menon, Brendan Van Rooyen, Cheng Soon Ong, Bob Williamson
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:125-134, 2015.

Abstract

Many supervised learning problems involve learning from samples whose labels are corrupted in some way. For example, each sample may have some constant probability of being incorrectly labelled (learning with label noise), or one may have a pool of unlabelled samples in lieu of negative samples (learning from positive and unlabelled data). This paper uses class-probability estimation to study these and other corruption processes belonging to the mutually contaminated distributions framework (Scott et al., 2013), with three conclusions. First, one can optimise balanced error and AUC without knowledge of the corruption process parameters. Second, given estimates of the corruption parameters, one can minimise a range of classification risks. Third, one can estimate the corruption parameters using only corrupted data. Experiments confirm the efficacy of class-probability estimation in learning from corrupted labels.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-menon15, title = {Learning from Corrupted Binary Labels via Class-Probability Estimation}, author = {Menon, Aditya and Rooyen, Brendan Van and Ong, Cheng Soon and Williamson, Bob}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {125--134}, 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/menon15.pdf}, url = { http://proceedings.mlr.press/v37/menon15.html }, abstract = {Many supervised learning problems involve learning from samples whose labels are corrupted in some way. For example, each sample may have some constant probability of being incorrectly labelled (learning with label noise), or one may have a pool of unlabelled samples in lieu of negative samples (learning from positive and unlabelled data). This paper uses class-probability estimation to study these and other corruption processes belonging to the mutually contaminated distributions framework (Scott et al., 2013), with three conclusions. First, one can optimise balanced error and AUC without knowledge of the corruption process parameters. Second, given estimates of the corruption parameters, one can minimise a range of classification risks. Third, one can estimate the corruption parameters using only corrupted data. Experiments confirm the efficacy of class-probability estimation in learning from corrupted labels.} }
Endnote
%0 Conference Paper %T Learning from Corrupted Binary Labels via Class-Probability Estimation %A Aditya Menon %A Brendan Van Rooyen %A Cheng Soon Ong %A Bob Williamson %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-menon15 %I PMLR %P 125--134 %U http://proceedings.mlr.press/v37/menon15.html %V 37 %X Many supervised learning problems involve learning from samples whose labels are corrupted in some way. For example, each sample may have some constant probability of being incorrectly labelled (learning with label noise), or one may have a pool of unlabelled samples in lieu of negative samples (learning from positive and unlabelled data). This paper uses class-probability estimation to study these and other corruption processes belonging to the mutually contaminated distributions framework (Scott et al., 2013), with three conclusions. First, one can optimise balanced error and AUC without knowledge of the corruption process parameters. Second, given estimates of the corruption parameters, one can minimise a range of classification risks. Third, one can estimate the corruption parameters using only corrupted data. Experiments confirm the efficacy of class-probability estimation in learning from corrupted labels.
RIS
TY - CPAPER TI - Learning from Corrupted Binary Labels via Class-Probability Estimation AU - Aditya Menon AU - Brendan Van Rooyen AU - Cheng Soon Ong AU - Bob Williamson BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-menon15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 125 EP - 134 L1 - http://proceedings.mlr.press/v37/menon15.pdf UR - http://proceedings.mlr.press/v37/menon15.html AB - Many supervised learning problems involve learning from samples whose labels are corrupted in some way. For example, each sample may have some constant probability of being incorrectly labelled (learning with label noise), or one may have a pool of unlabelled samples in lieu of negative samples (learning from positive and unlabelled data). This paper uses class-probability estimation to study these and other corruption processes belonging to the mutually contaminated distributions framework (Scott et al., 2013), with three conclusions. First, one can optimise balanced error and AUC without knowledge of the corruption process parameters. Second, given estimates of the corruption parameters, one can minimise a range of classification risks. Third, one can estimate the corruption parameters using only corrupted data. Experiments confirm the efficacy of class-probability estimation in learning from corrupted labels. ER -
APA
Menon, A., Rooyen, B.V., Ong, C.S. & Williamson, B.. (2015). Learning from Corrupted Binary Labels via Class-Probability Estimation. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:125-134 Available from http://proceedings.mlr.press/v37/menon15.html .

Related Material