Details of Research Outputs

TitleLearning to Re-weight Examples with Optimal Transport for Imbalanced Classification
Author (Name in English or Pinyin)
Guo, Dandan1,2; Li, Zhuo3,4; Zheng, Meixi5; Zhao, He6; Zhou, Mingyuan7; Zha, Hongyuan1,8
Date Issued2022
Conference Name36th Conference on Neural Information Processing Systems, NeurIPS 2022
Source PublicationAdvances in Neural Information Processing Systems
Conference DateNovember 28, 2022 - December 9, 2022
Conference PlaceNew Orleans, LA, United states
PublisherNeural information processing systems foundation
AbstractImbalanced data pose challenges for deep learning based classification models. One of the most widely-used approaches for tackling imbalanced data is re-weighting, where training samples are associated with different weights in the loss function. Most of existing re-weighting approaches treat the example weights as the learnable parameter and optimize the weights on the meta set, entailing expensive bilevel optimization. In this paper, we propose a novel re-weighting method based on optimal transport (OT) from a distributional point of view. Specifically, we view the training set as an imbalanced distribution over its samples, which is transported by OT to a balanced distribution obtained from the meta set. The weights of the training samples are the probability mass of the imbalanced distribution and learned by minimizing the OT distance between the two distributions. Compared with existing methods, our proposed one disengages the dependence of the weight learning on the concerned classifier at each iteration. Experiments on image, text and point cloud datasets demonstrate that our proposed re-weighting method has excellent performance, achieving state-of-the-art results in many cases and providing a promising tool for addressing the imbalanced classification issue. The code has been made available at © 2022 Neural information processing systems foundation. All rights reserved.
KeywordClassification (of information) Deep learning Probability distributions Sampling Text processing Bi-level optimization Classification models Imbalanced classification Imbalanced data Loss functions Optimal transport Re-weighting Training sample Weighting approaches Weighting methods
Indexed ByEI
EI Accession Number20232614294820
EI Classification Number461.4 Ergonomics and Human Factors Engineering - 716.1 Information Theory and Signal Processing - 903.1 Information Sources and Analysis - 903.3 Information Retrieval and Use - 921.6 Numerical Methods - 922.1 Probability Theory
Original Document TypeConference article (CA)
Published range国外学术期刊
Volume Issue Pagesv 35,
Data SourceEI
Document TypeConference paper
CollectionSchool of Data Science
Institute of Robotics and Intelligent Manufacturing
1.School of Data Science, The Chinese University of HongKong, Shenzhen, China
2.Institute of Robotics and Intelligent Manufacturing, China
3.School of Science and Engineering, The Chinese University of HongKong, Shenzhen, China
4.Shenzhen Research Institute of Big Data, China
5.Xidian University, China
6.CSIRO's Data61, Australia
7.The University of Texas, Austin, United States
8.Shenzhen Institute of Artificial Intelligence and Robotics for Society, China
First Author AffilicationSchool of Data Science;  Institute of Robotics and Intelligent Manufacturing
Recommended Citation
GB/T 7714
Guo, Dandan,Li, Zhuo,Zheng, Meixiet al. Learning to Re-weight Examples with Optimal Transport for Imbalanced Classification[C]:Neural information processing systems foundation,2022.
Files in This Item:
There are no files associated with this item.
Related Services
Usage statistics
Google Scholar
Similar articles in Google Scholar
[Guo, Dandan]'s Articles
[Li, Zhuo]'s Articles
[Zheng, Meixi]'s Articles
Baidu academic
Similar articles in Baidu academic
[Guo, Dandan]'s Articles
[Li, Zhuo]'s Articles
[Zheng, Meixi]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Guo, Dandan]'s Articles
[Li, Zhuo]'s Articles
[Zheng, Meixi]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.