Making deep neural networks robust to label noise: A reweighting loss and data filtration

Zhengwen Zhang*, Yan Li, Yunjie Li, Ying Qin

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

Deep neural networks (DNNs) have achieved astonishing results on a variety of supervised learning tasks owing to a large scale of expert-labeled datasets. However, as recent researches pointed out, the generalization performance of DNNs deteriorates rapidly when training data contains label noise. To alleviate the problem of easily overfitting to label-corrupted data when training DNNs, this paper proposes a robust loss function by reweighting the standard Cross-Entropy loss. For obtaining more robust DNNs under label noise, we further design a framework to jointly optimize model parameters and filtering noisy data during training. Our methods can be thus viewed as online curriculum learning based on both loss function and training datasets. A great deal of experiments have conducted on CIFAR-10, CIFAR-100 and ImageNet under two types of label noise. The results of our proposed method outperforms other state-of-the-art methods.

Original languageEnglish
Title of host publication2019 IEEE 4th International Conference on Signal and Image Processing, ICSIP 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages289-293
Number of pages5
ISBN (Electronic)9781728136608
DOIs
Publication statusPublished - Jul 2019
Event4th IEEE International Conference on Signal and Image Processing, ICSIP 2019 - Wuxi, China
Duration: 19 Jul 201921 Jul 2019

Publication series

Name2019 IEEE 4th International Conference on Signal and Image Processing, ICSIP 2019

Conference

Conference4th IEEE International Conference on Signal and Image Processing, ICSIP 2019
Country/TerritoryChina
CityWuxi
Period19/07/1921/07/19

Keywords

  • Curriculum learning
  • Data filtration
  • Deep neural networks
  • Label noise
  • Robust loss function

Fingerprint

Dive into the research topics of 'Making deep neural networks robust to label noise: A reweighting loss and data filtration'. Together they form a unique fingerprint.

Cite this