TY - GEN
T1 - Provably End-to-end Label-noise Learning without Anchor Points
AU - Li, Xuefeng
AU - Liu, Tongliang
AU - Han, Bo
AU - Niu, Gang
AU - Sugiyama, Masashi
N1 - Funding Information:
XL was supported by an Australian Government RTP Scholarship. TL was supported by Australian Research Council Project DE-190101473. BH was supported by the RGC Early Career Scheme No. 22200720, NSFC Young Scientists Fund No. 62006202 and HKBU CSD Departmental Incentive Grant. GN and MS were supported by JST AIP Acceleration Research Grant Number JPMJCR20U3, Japan. MS was also supported by Institute for AI and Beyond, UTokyo. Authors also thank for the help from Dr. Alan Blair, Kevin Lam, Yivan Zhang and members of the Trustworthy Machine Learning Lab at the University of Sydney.
Publisher Copyright:
Copyright © 2021 by the author(s)
PY - 2021
Y1 - 2021
N2 - In label-noise learning, the transition matrix plays a key role in building statistically consistent classifiers. Existing consistent estimators for the transition matrix have been developed by exploiting anchor points. However, the anchor-point assumption is not always satisfied in real scenarios. In this paper, we propose an end-to-end framework for solving label-noise learning without anchor points, in which we simultaneously optimize two objectives: the cross entropy loss between the noisy label and the predicted probability by the neural network, and the volume of the simplex formed by the columns of the transition matrix. Our proposed framework can identify the transition matrix if the clean class-posterior probabilities are sufficiently scattered. This is by far the mildest assumption under which the transition matrix is provably identifiable and the learned classifier is statistically consistent. Experimental results on benchmark datasets demonstrate the effectiveness and robustness of the proposed method.
AB - In label-noise learning, the transition matrix plays a key role in building statistically consistent classifiers. Existing consistent estimators for the transition matrix have been developed by exploiting anchor points. However, the anchor-point assumption is not always satisfied in real scenarios. In this paper, we propose an end-to-end framework for solving label-noise learning without anchor points, in which we simultaneously optimize two objectives: the cross entropy loss between the noisy label and the predicted probability by the neural network, and the volume of the simplex formed by the columns of the transition matrix. Our proposed framework can identify the transition matrix if the clean class-posterior probabilities are sufficiently scattered. This is by far the mildest assumption under which the transition matrix is provably identifiable and the learned classifier is statistically consistent. Experimental results on benchmark datasets demonstrate the effectiveness and robustness of the proposed method.
UR - http://www.scopus.com/inward/record.url?scp=85137948201&partnerID=8YFLogxK
UR - https://www.proceedings.com/63018.html
M3 - Conference proceeding
AN - SCOPUS:85137948201
T3 - Proceedings of Machine Learning Research
SP - 6403
EP - 6413
BT - Proceedings of the 38th International Conference on Machine Learning, ICML 2021
PB - Mathematical Research Press
T2 - 38th International Conference on Machine Learning, ICML 2021
Y2 - 18 July 2021 through 24 July 2021
ER -