logo
科言猫学术研究的AI总结
language
手机端手机端
  • Phone Feature微信小程序
简体中文简体中文
notice消息通知
博客博客
delete
arrow
返回

Confidence-Based PU Learning With Instance-Dependent Label Noise

delete2025-08-01
delete0
PRE
AI
X
Xijia Tang
C
Chao Xu
H
Hong Tao
马潇宇 封面图
马潇宇 (Xiaoyu Ma)
C
Chenping Hou
DOI：10.1109/TNNLS.2025.3549510delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
正类和未标记（PU）学习，即仅使用PU数据训练二元分类器的方法，近年来已引起广泛关注。传统的PU学习通常假设所有正样本都被准确标注。然而，由于样本模糊性和算法不足等原因，在此场景中标签噪声几乎不可避免。当前PU算法忽略了正集中的标签噪声问题，而该噪声在实际应用中往往偏向特定实例而非均匀分布。我们将这一重要但研究不足的问题定义为具有实例依赖标签噪声的PU学习（PUIDN）。为消除IDN的负面影响，我们利用正集中每个实例的置信度分数，在不假设噪声分布的情况下建立样本与标签之间的联系。随后，我们提出了一种考虑标签和置信度信息的无偏分类风险估计器，该估计器可基于PUIDN数据及其置信度分数直接计算。此外，我们的分类框架整合了一种基于不同置信度信息相关性的交替迭代优化策略，从而减轻了对训练数据的额外需求。理论上，我们推导了所提方法的经验风险界。实验上，通过多种类型的数值结果验证了该方法的有效性。
Keyword:
Biased label
confidence information
instance-dependent label noise (IND)
positive and unlabeled (PU) learning

期刊

IEEE Transactions on Neural Networks and Learning Systems 封面图
IEEE Transactions on Neural Networks and Learning Systems
IF:
8.9
论文数:
7.6K
被引数:
7.2W

机构

N
National University of Defense Technology
学者数:
3.3K
论文数: 1.0K
被引数: 8.2K
引用论文

引用论文

A Robust AUC Maximization Framework With Simultaneous Outlier Detection and Feature Selection for Positive-Unlabeled Classification
errIEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
IF8.9
err2019-10-01
err21
errOAAI
errRen, Ke; Yang, Haichuan; Zhao, Yu; Chen, Wu; Xue, Mingshan; Miao, Hongyu; Huang, Shuai; Liu, Ji
err分享
err收藏
Attribute augmentation-based label integration for crowdsourcing
errFRONTIERS OF COMPUTER SCIENCE
IF4.6
err2022-12-24
err26
PREAI
errZhang, Yao; Jiang, Liangxiao; Li, Chaoqun
err分享
err收藏
Centroid Estimation With Guaranteed Efficiency: A General Framework for Weakly Supervised Learning
errIEEE Transactions on Pattern Analysis and Machine Intelligence
IF0
err2022-06-01
err0
PREAI
errChen Gong; Jian Yang; Jane You; Masashi Sugiyama
err分享
err收藏
Positive and Unlabeled Learning via Loss Decomposition and Centroid Estimation
errProceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence
IF0
err2018-07-01
err0
errOAAI
errHong Shi; Shaojun Pan; Jian Yang; Chen Gong
err分享
err收藏
A brief introduction to weakly supervised learning弱监督学习简介
errNATIONAL SCIENCE REVIEW
IF17.1
err2017-08-25
err1.3K
errOAAI
errZhou, Zhi-Hua
err分享
err收藏
Probability in Banach Spaces
err
IF0
err1991-01-01
err0
PREAI
errMichel Ledoux; Michel Talagrand
err分享
err收藏
Data fusing and joint training for learning with noisy labels
errFRONTIERS OF COMPUTER SCIENCE
IF4.6
err2022-04-02
err10
PREAI
errWei, Yi; Xue, Mei; Liu, Xin; Xu, Pengxiang
err分享
err收藏
A bagging SVM to learn from positive and unlabeled examples
errPATTERN RECOGNITION LETTERS
IF3.3
err2014-02-01
err233
errOAAI
errMordelet, F.; Vert, J. -P.
err分享
err收藏
Learning classifiers from only positive and unlabeled data
errProceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining
IF0
err2008-08-24
err0
errOAAI
errCharles Elkan; Keith Noto
err分享
err收藏
Incorporating Semi-Supervised and Positive-Unlabeled Learning for Boosting Full Reference Image Quality Assessment
err2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2022-06-01
err0
errOAAI
errYue Cao; Zhaolin Wan; Dongwei Ren; Zifei Yan; Wangmeng Zuo
err分享
err收藏
学者 查看更多内容