Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels

Lu Jiang, Di Huang, Mason Liu, Weilong Yang
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:4804-4815, 2020.

Abstract

Performing controlled experiments on noisy data is essential in understanding deep learning across noise levels. Due to the lack of suitable datasets, previous research has only examined deep learning on controlled synthetic label noise, and real-world label noise has never been studied in a controlled setting. This paper makes three contributions. First, we establish the first benchmark of controlled real-world label noise from the web. This new benchmark enables us to study the web label noise in a controlled setting for the first time. The second contribution is a simple but effective method to overcome both synthetic and real noisy labels. We show that our method achieves the best result on our dataset as well as on two public benchmarks (CIFAR and WebVision). Third, we conduct the largest study by far into understanding deep neural networks trained on noisy labels across different noise levels, noise types, network architectures, and training settings.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-jiang20c, title = {Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels}, author = {Jiang, Lu and Huang, Di and Liu, Mason and Yang, Weilong}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {4804--4815}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/jiang20c/jiang20c.pdf}, url = {http://proceedings.mlr.press/v119/jiang20c.html}, abstract = {Performing controlled experiments on noisy data is essential in understanding deep learning across noise levels. Due to the lack of suitable datasets, previous research has only examined deep learning on controlled synthetic label noise, and real-world label noise has never been studied in a controlled setting. This paper makes three contributions. First, we establish the first benchmark of controlled real-world label noise from the web. This new benchmark enables us to study the web label noise in a controlled setting for the first time. The second contribution is a simple but effective method to overcome both synthetic and real noisy labels. We show that our method achieves the best result on our dataset as well as on two public benchmarks (CIFAR and WebVision). Third, we conduct the largest study by far into understanding deep neural networks trained on noisy labels across different noise levels, noise types, network architectures, and training settings.} }
Endnote
%0 Conference Paper %T Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels %A Lu Jiang %A Di Huang %A Mason Liu %A Weilong Yang %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-jiang20c %I PMLR %P 4804--4815 %U http://proceedings.mlr.press/v119/jiang20c.html %V 119 %X Performing controlled experiments on noisy data is essential in understanding deep learning across noise levels. Due to the lack of suitable datasets, previous research has only examined deep learning on controlled synthetic label noise, and real-world label noise has never been studied in a controlled setting. This paper makes three contributions. First, we establish the first benchmark of controlled real-world label noise from the web. This new benchmark enables us to study the web label noise in a controlled setting for the first time. The second contribution is a simple but effective method to overcome both synthetic and real noisy labels. We show that our method achieves the best result on our dataset as well as on two public benchmarks (CIFAR and WebVision). Third, we conduct the largest study by far into understanding deep neural networks trained on noisy labels across different noise levels, noise types, network architectures, and training settings.
APA
Jiang, L., Huang, D., Liu, M. & Yang, W.. (2020). Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:4804-4815 Available from http://proceedings.mlr.press/v119/jiang20c.html.

Related Material