Searching to Exploit Memorization Effect in Learning with Noisy Labels

Quanming Yao, Hansi Yang, Bo Han, Gang Niu, James Tin-Yau Kwok
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:10789-10798, 2020.

Abstract

Sample selection approaches are popular in robust learning from noisy labels. However, how to properly control the selection process so that deep networks can benefit from the memorization effect is a hard problem. In this paper, motivated by the success of automated machine learning (AutoML), we model this issue as a function approximation problem. Specifically, we design a domain-specific search space based on general patterns of the memorization effect and propose a novel Newton algorithm to solve the bi-level optimization problem efficiently. We further provide a theoretical analysis of the algorithm, which ensures a good approximation to critical points. Experiments are performed on both benchmark and real-world data sets. Results demonstrate that the proposed method is much better than the state-of-the-art noisy-label-learning approaches, and also much more efficient than existing AutoML algorithms.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-yao20b, title = {Searching to Exploit Memorization Effect in Learning with Noisy Labels}, author = {Yao, Quanming and Yang, Hansi and Han, Bo and Niu, Gang and Kwok, James Tin-Yau}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {10789--10798}, year = {2020}, editor = {Hal Daumé III and Aarti Singh}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/yao20b/yao20b.pdf}, url = { http://proceedings.mlr.press/v119/yao20b.html }, abstract = {Sample selection approaches are popular in robust learning from noisy labels. However, how to properly control the selection process so that deep networks can benefit from the memorization effect is a hard problem. In this paper, motivated by the success of automated machine learning (AutoML), we model this issue as a function approximation problem. Specifically, we design a domain-specific search space based on general patterns of the memorization effect and propose a novel Newton algorithm to solve the bi-level optimization problem efficiently. We further provide a theoretical analysis of the algorithm, which ensures a good approximation to critical points. Experiments are performed on both benchmark and real-world data sets. Results demonstrate that the proposed method is much better than the state-of-the-art noisy-label-learning approaches, and also much more efficient than existing AutoML algorithms.} }
Endnote
%0 Conference Paper %T Searching to Exploit Memorization Effect in Learning with Noisy Labels %A Quanming Yao %A Hansi Yang %A Bo Han %A Gang Niu %A James Tin-Yau Kwok %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-yao20b %I PMLR %P 10789--10798 %U http://proceedings.mlr.press/v119/yao20b.html %V 119 %X Sample selection approaches are popular in robust learning from noisy labels. However, how to properly control the selection process so that deep networks can benefit from the memorization effect is a hard problem. In this paper, motivated by the success of automated machine learning (AutoML), we model this issue as a function approximation problem. Specifically, we design a domain-specific search space based on general patterns of the memorization effect and propose a novel Newton algorithm to solve the bi-level optimization problem efficiently. We further provide a theoretical analysis of the algorithm, which ensures a good approximation to critical points. Experiments are performed on both benchmark and real-world data sets. Results demonstrate that the proposed method is much better than the state-of-the-art noisy-label-learning approaches, and also much more efficient than existing AutoML algorithms.
APA
Yao, Q., Yang, H., Han, B., Niu, G. & Kwok, J.T.. (2020). Searching to Exploit Memorization Effect in Learning with Noisy Labels. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:10789-10798 Available from http://proceedings.mlr.press/v119/yao20b.html .

Related Material