Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions

Samples in large-scale datasets may be mislabeled for various reasons, and deep models are inclined to over-fit some noisy samples using conventional training procedures. The key solution is to alleviate the harm of these noisy labels. Many existing methods try to divide training data into clean and...

Full description

Saved in:
Bibliographic Details
Main Authors: Lida Yu, Xuefeng Liang, Chang Cao, Longshan Yao, Xingyu Liu
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/11/3369
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849330925497745408
author Lida Yu
Xuefeng Liang
Chang Cao
Longshan Yao
Xingyu Liu
author_facet Lida Yu
Xuefeng Liang
Chang Cao
Longshan Yao
Xingyu Liu
author_sort Lida Yu
collection DOAJ
description Samples in large-scale datasets may be mislabeled for various reasons, and deep models are inclined to over-fit some noisy samples using conventional training procedures. The key solution is to alleviate the harm of these noisy labels. Many existing methods try to divide training data into clean and noisy subsets in terms of loss values. We observe that a reason hindering the better performance of deep models is the uncertain samples, which have relatively small losses and often appear in real-world datasets. Due to small losses, many uncertain noisy samples are divided into the clean subset and then degrade models’ performance. Instead, we propose a Tripartite solution to partition training data into three subsets, <i>uncertain</i>, <i>clean</i> and <i>noisy</i> according to the following criteria: the inconsistency of the predictions of two networks and the given labels. Tripartite considerably improves the quality of the clean subset. Moreover, to maximize the value of clean samples in the uncertain subset and minimize the harm of noisy labels, we apply low-weight learning and a semi-supervised learning, respectively. Extensive experiments demonstrate that Tripartite can filter out noisy samples more precisely and outperforms most state-of-the-art methods on four benchmark datasets and especially real-world datasets.
format Article
id doaj-art-fbe5b4ebaece48cab6fc3f9a8ee5dd3c
institution Kabale University
issn 1424-8220
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-fbe5b4ebaece48cab6fc3f9a8ee5dd3c2025-08-20T03:46:46ZengMDPI AGSensors1424-82202025-05-012511336910.3390/s25113369Tripartite: Tackling Realistic Noisy Labels with More Precise PartitionsLida Yu0Xuefeng Liang1Chang Cao2Longshan Yao3Xingyu Liu4School of Arts and Sciences, Beijing Normal University, Beijing 100875, ChinaSchool of Artificial Intelligence, Xidian University, Xi’an 710071, ChinaGuangzhou Institute of Technology, Xidian University, Guangzhou 510555, ChinaSchool of Artificial Intelligence, Xidian University, Xi’an 710071, ChinaGuangzhou Institute of Technology, Xidian University, Guangzhou 510555, ChinaSamples in large-scale datasets may be mislabeled for various reasons, and deep models are inclined to over-fit some noisy samples using conventional training procedures. The key solution is to alleviate the harm of these noisy labels. Many existing methods try to divide training data into clean and noisy subsets in terms of loss values. We observe that a reason hindering the better performance of deep models is the uncertain samples, which have relatively small losses and often appear in real-world datasets. Due to small losses, many uncertain noisy samples are divided into the clean subset and then degrade models’ performance. Instead, we propose a Tripartite solution to partition training data into three subsets, <i>uncertain</i>, <i>clean</i> and <i>noisy</i> according to the following criteria: the inconsistency of the predictions of two networks and the given labels. Tripartite considerably improves the quality of the clean subset. Moreover, to maximize the value of clean samples in the uncertain subset and minimize the harm of noisy labels, we apply low-weight learning and a semi-supervised learning, respectively. Extensive experiments demonstrate that Tripartite can filter out noisy samples more precisely and outperforms most state-of-the-art methods on four benchmark datasets and especially real-world datasets.https://www.mdpi.com/1424-8220/25/11/3369realistic noisy labeluncertain samplestripartitionbipartition
spellingShingle Lida Yu
Xuefeng Liang
Chang Cao
Longshan Yao
Xingyu Liu
Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
Sensors
realistic noisy label
uncertain samples
tripartition
bipartition
title Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
title_full Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
title_fullStr Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
title_full_unstemmed Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
title_short Tripartite: Tackling Realistic Noisy Labels with More Precise Partitions
title_sort tripartite tackling realistic noisy labels with more precise partitions
topic realistic noisy label
uncertain samples
tripartition
bipartition
url https://www.mdpi.com/1424-8220/25/11/3369
work_keys_str_mv AT lidayu tripartitetacklingrealisticnoisylabelswithmoreprecisepartitions
AT xuefengliang tripartitetacklingrealisticnoisylabelswithmoreprecisepartitions
AT changcao tripartitetacklingrealisticnoisylabelswithmoreprecisepartitions
AT longshanyao tripartitetacklingrealisticnoisylabelswithmoreprecisepartitions
AT xingyuliu tripartitetacklingrealisticnoisylabelswithmoreprecisepartitions