MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels

Lu Jiang, Zhengyuan Zhou, Thomas Leung, Li-Jia Li, Li Fei-Fei
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:2304-2313, 2018.

Abstract

Recent deep networks are capable of memorizing the entire data even when the labels are completely random. To overcome the overfitting on corrupted labels, we propose a novel technique of learning another neural network, called MentorNet, to supervise the training of the base deep networks, namely, StudentNet. During training, MentorNet provides a curriculum (sample weighting scheme) for StudentNet to focus on the sample the label of which is probably correct. Unlike the existing curriculum that is usually predefined by human experts, MentorNet learns a data-driven curriculum dynamically with StudentNet. Experimental results demonstrate that our approach can significantly improve the generalization performance of deep networks trained on corrupted training data. Notably, to the best of our knowledge, we achieve the best-published result on WebVision, a large benchmark containing 2.2 million images of real-world noisy labels.

Cite this Paper


BibTeX
@InProceedings{pmlr-v80-jiang18c, title = {{M}entor{N}et: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels}, author = {Jiang, Lu and Zhou, Zhengyuan and Leung, Thomas and Li, Li-Jia and Fei-Fei, Li}, booktitle = {Proceedings of the 35th International Conference on Machine Learning}, pages = {2304--2313}, year = {2018}, editor = {Dy, Jennifer and Krause, Andreas}, volume = {80}, series = {Proceedings of Machine Learning Research}, month = {10--15 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v80/jiang18c/jiang18c.pdf}, url = {https://proceedings.mlr.press/v80/jiang18c.html}, abstract = {Recent deep networks are capable of memorizing the entire data even when the labels are completely random. To overcome the overfitting on corrupted labels, we propose a novel technique of learning another neural network, called MentorNet, to supervise the training of the base deep networks, namely, StudentNet. During training, MentorNet provides a curriculum (sample weighting scheme) for StudentNet to focus on the sample the label of which is probably correct. Unlike the existing curriculum that is usually predefined by human experts, MentorNet learns a data-driven curriculum dynamically with StudentNet. Experimental results demonstrate that our approach can significantly improve the generalization performance of deep networks trained on corrupted training data. Notably, to the best of our knowledge, we achieve the best-published result on WebVision, a large benchmark containing 2.2 million images of real-world noisy labels.} }
Endnote
%0 Conference Paper %T MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels %A Lu Jiang %A Zhengyuan Zhou %A Thomas Leung %A Li-Jia Li %A Li Fei-Fei %B Proceedings of the 35th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jennifer Dy %E Andreas Krause %F pmlr-v80-jiang18c %I PMLR %P 2304--2313 %U https://proceedings.mlr.press/v80/jiang18c.html %V 80 %X Recent deep networks are capable of memorizing the entire data even when the labels are completely random. To overcome the overfitting on corrupted labels, we propose a novel technique of learning another neural network, called MentorNet, to supervise the training of the base deep networks, namely, StudentNet. During training, MentorNet provides a curriculum (sample weighting scheme) for StudentNet to focus on the sample the label of which is probably correct. Unlike the existing curriculum that is usually predefined by human experts, MentorNet learns a data-driven curriculum dynamically with StudentNet. Experimental results demonstrate that our approach can significantly improve the generalization performance of deep networks trained on corrupted training data. Notably, to the best of our knowledge, we achieve the best-published result on WebVision, a large benchmark containing 2.2 million images of real-world noisy labels.
APA
Jiang, L., Zhou, Z., Leung, T., Li, L. & Fei-Fei, L.. (2018). MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels. Proceedings of the 35th International Conference on Machine Learning, in Proceedings of Machine Learning Research 80:2304-2313 Available from https://proceedings.mlr.press/v80/jiang18c.html.

Related Material