Beyond Class-Conditional Assumption: A Primary Attempt to Combat Instance-Dependent Label Noise

被引:0
|
作者
Chen, Pengfei [1 ]
Ye, Junjie [2 ]
Chen, Guangyong [3 ]
Zhao, Jingwei [2 ]
Heng, Pheng-Ann [1 ,3 ]
机构
[1] Chinese Univ Hong Kong, Hong Kong, Peoples R China
[2] VIVO AI Lab, Shenzhen, Peoples R China
[3] Chinese Acad Sci, Shenzhen Inst Adv Technol, Shenzhen Key Lab Virtual Real & Human Interact Te, Shenzhen, Peoples R China
来源
THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE | 2021年 / 35卷
基金
中国国家自然科学基金;
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Supervised learning under label noise has seen numerous advances recently, while existing theoretical findings and empirical results broadly build up on the class-conditional noise (CCN) assumption that the noise is independent of input features given the true label. In this work, we present a theoretical hypothesis testing and prove that noise in real-world dataset is unlikely to be CCN, which confirms that label noise should depend on the instance and justifies the urgent need to go beyond the CCN assumption.The theoretical results motivate us to study the more general and practical-relevant instance-dependent noise (IDN). To stimulate the development of theory and methodology on IDN, we formalize an algorithm to generate controllable IDN and present both theoretical and empirical evidence to show that IDN is semantically meaningful and challenging. As a primary attempt to combat IDN, we present a tiny algorithm termed self-evolution average label (SEAL), which not only stands out under IDN with various noise fractions, but also improves the generalization on real-world noise benchmark Clothing1M. Our code is released. Notably, our theoretical analysis in Section 2 provides rigorous motivations for studying IDN, which is an important topic that deserves more research attention in future.
引用
收藏
页码:11442 / 11450
页数:9
相关论文
共 22 条
  • [21] Instance-Dependent Noisy-Label Learning with Graphical Model Based Noise-Rate Estimation
    Garg, Arpit
    Cuong Nguyen
    Felix, Rafael
    Thanh-Toan Do
    Carneiro, Gustavo
    COMPUTER VISION-ECCV 2024, PT IV, 2025, 15062 : 372 - 389
  • [22] Typicality- and instance-dependent label noise-combating: a novel framework for simulating and combating real-world noisy labels for endoscopic polyp classification
    Gao, Yun
    Fu, Junhu
    Wang, Yuanyuan
    Guo, Yi
    VISUAL COMPUTING FOR INDUSTRY BIOMEDICINE AND ART, 2024, 7 (01)