Robust Deep Ordinal Regression under Label Noise

Bhanu Garg, Naresh Manwani
; Proceedings of The 12th Asian Conference on Machine Learning, PMLR 129:782-796, 2020.

Abstract

The real-world data is often susceptible to label noise, which might constrict the effectiveness of the existing state of the art algorithms for ordinal regression. Existing works on ordinal regression do not take label noise into account. We propose a theoretically grounded approach for class conditional label noise in ordinal regression problems. We present a deep learning implementation of two commonly used loss functions for ordinal regression that is both - 1) robust to label noise, and 2) rank consistent for a good ranking rule. We verify these properties of the algorithm empirically and show robustness to label noise on real data and rank consistency. To the best of our knowledge, this is the first approach for robust ordinal regression models.

Cite this Paper


BibTeX
@InProceedings{pmlr-v129-garg20a, title = {Robust Deep Ordinal Regression under Label Noise}, author = {Garg, Bhanu and Manwani, Naresh}, booktitle = {Proceedings of The 12th Asian Conference on Machine Learning}, pages = {782--796}, year = {2020}, editor = {Sinno Jialin Pan and Masashi Sugiyama}, volume = {129}, series = {Proceedings of Machine Learning Research}, address = {Bangkok, Thailand}, month = {18--20 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v129/garg20a/garg20a.pdf}, url = {http://proceedings.mlr.press/v129/garg20a.html}, abstract = {The real-world data is often susceptible to label noise, which might constrict the effectiveness of the existing state of the art algorithms for ordinal regression. Existing works on ordinal regression do not take label noise into account. We propose a theoretically grounded approach for class conditional label noise in ordinal regression problems. We present a deep learning implementation of two commonly used loss functions for ordinal regression that is both - 1) robust to label noise, and 2) rank consistent for a good ranking rule. We verify these properties of the algorithm empirically and show robustness to label noise on real data and rank consistency. To the best of our knowledge, this is the first approach for robust ordinal regression models.} }
Endnote
%0 Conference Paper %T Robust Deep Ordinal Regression under Label Noise %A Bhanu Garg %A Naresh Manwani %B Proceedings of The 12th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Sinno Jialin Pan %E Masashi Sugiyama %F pmlr-v129-garg20a %I PMLR %J Proceedings of Machine Learning Research %P 782--796 %U http://proceedings.mlr.press %V 129 %W PMLR %X The real-world data is often susceptible to label noise, which might constrict the effectiveness of the existing state of the art algorithms for ordinal regression. Existing works on ordinal regression do not take label noise into account. We propose a theoretically grounded approach for class conditional label noise in ordinal regression problems. We present a deep learning implementation of two commonly used loss functions for ordinal regression that is both - 1) robust to label noise, and 2) rank consistent for a good ranking rule. We verify these properties of the algorithm empirically and show robustness to label noise on real data and rank consistency. To the best of our knowledge, this is the first approach for robust ordinal regression models.
APA
Garg, B. & Manwani, N.. (2020). Robust Deep Ordinal Regression under Label Noise. Proceedings of The 12th Asian Conference on Machine Learning, in PMLR 129:782-796

Related Material