GenLabel: Mixup Relabeling using Generative Models

Jy-Yong Sohn, Liang Shang, Hongxu Chen, Jaekyun Moon, Dimitris Papailiopoulos, Kangwook Lee
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:20278-20313, 2022.

Abstract

Mixup is a data augmentation method that generates new data points by mixing a pair of input data. While mixup generally improves the prediction performance, it sometimes degrades the performance. In this paper, we first identify the main causes of this phenomenon by theoretically and empirically analyzing the mixup algorithm. To resolve this, we propose GenLabel, a simple yet effective relabeling algorithm designed for mixup. In particular, GenLabel helps the mixup algorithm correctly label mixup samples by learning the class-conditional data distribution using generative models. Via theoretical and empirical analysis, we show that mixup, when used together with GenLabel, can effectively resolve the aforementioned phenomenon, improving the accuracy of mixup-trained model.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-sohn22a, title = {{G}en{L}abel: Mixup Relabeling using Generative Models}, author = {Sohn, Jy-Yong and Shang, Liang and Chen, Hongxu and Moon, Jaekyun and Papailiopoulos, Dimitris and Lee, Kangwook}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {20278--20313}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/sohn22a/sohn22a.pdf}, url = {https://proceedings.mlr.press/v162/sohn22a.html}, abstract = {Mixup is a data augmentation method that generates new data points by mixing a pair of input data. While mixup generally improves the prediction performance, it sometimes degrades the performance. In this paper, we first identify the main causes of this phenomenon by theoretically and empirically analyzing the mixup algorithm. To resolve this, we propose GenLabel, a simple yet effective relabeling algorithm designed for mixup. In particular, GenLabel helps the mixup algorithm correctly label mixup samples by learning the class-conditional data distribution using generative models. Via theoretical and empirical analysis, we show that mixup, when used together with GenLabel, can effectively resolve the aforementioned phenomenon, improving the accuracy of mixup-trained model.} }
Endnote
%0 Conference Paper %T GenLabel: Mixup Relabeling using Generative Models %A Jy-Yong Sohn %A Liang Shang %A Hongxu Chen %A Jaekyun Moon %A Dimitris Papailiopoulos %A Kangwook Lee %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-sohn22a %I PMLR %P 20278--20313 %U https://proceedings.mlr.press/v162/sohn22a.html %V 162 %X Mixup is a data augmentation method that generates new data points by mixing a pair of input data. While mixup generally improves the prediction performance, it sometimes degrades the performance. In this paper, we first identify the main causes of this phenomenon by theoretically and empirically analyzing the mixup algorithm. To resolve this, we propose GenLabel, a simple yet effective relabeling algorithm designed for mixup. In particular, GenLabel helps the mixup algorithm correctly label mixup samples by learning the class-conditional data distribution using generative models. Via theoretical and empirical analysis, we show that mixup, when used together with GenLabel, can effectively resolve the aforementioned phenomenon, improving the accuracy of mixup-trained model.
APA
Sohn, J., Shang, L., Chen, H., Moon, J., Papailiopoulos, D. & Lee, K.. (2022). GenLabel: Mixup Relabeling using Generative Models. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:20278-20313 Available from https://proceedings.mlr.press/v162/sohn22a.html.

Related Material