Learning from Corrupted Binary Labels via Class-Probability Estimation
Menon, Aditya; Van Rooyen, Brendan; Ong, Cheng Song; Williamson, Robert
Description
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...[Show more]
dc.contributor.author | Menon, Aditya | |
---|---|---|
dc.contributor.author | Van Rooyen, Brendan | |
dc.contributor.author | Ong, Cheng Song | |
dc.contributor.author | Williamson, Robert | |
dc.coverage.spatial | Lille, France | |
dc.date.accessioned | 2016-06-14T23:21:14Z | |
dc.date.created | July 6-11 2015 | |
dc.identifier.uri | http://hdl.handle.net/1885/103788 | |
dc.description.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 | |
dc.publisher | JMLR | |
dc.relation.ispartofseries | International Conference on Machine Learning 2015 ICML | |
dc.title | Learning from Corrupted Binary Labels via Class-Probability Estimation | |
dc.type | Conference paper | |
local.description.notes | Imported from ARIES | |
local.description.refereed | Yes | |
dc.date.issued | 2015 | |
local.identifier.absfor | 080107 - Natural Language Processing | |
local.identifier.ariespublication | u4334215xPUB1514 | |
local.type.status | Published Version | |
local.contributor.affiliation | Menon, Aditya, College of Engineering and Computer Science, ANU | |
local.contributor.affiliation | Van Rooyen, Brendan, College of Engineering and Computer Science, ANU | |
local.contributor.affiliation | Ong, Cheng Song, College of Engineering and Computer Science, ANU | |
local.contributor.affiliation | Williamson, Robert, College of Engineering and Computer Science, ANU | |
local.description.embargo | 2037-12-31 | |
local.bibliographicCitation.startpage | 125 | |
local.bibliographicCitation.lastpage | 134 | |
local.identifier.absseo | 970108 - Expanding Knowledge in the Information and Computing Sciences | |
dc.date.updated | 2016-06-14T09:03:20Z | |
Collections | ANU Research Publications |
Download
File | Description | Size | Format | Image |
---|---|---|---|---|
01_Menon_Learning_from_Corrupted_Binary_2015.pdf | 342.01 kB | Adobe PDF | Request a copy |
Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.
Updated: 17 November 2022/ Responsible Officer: University Librarian/ Page Contact: Library Systems & Web Coordinator