Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning

In training-based Machine Learning applications, the training data are frequently labeled by non-experts and expose substantial label noise which greatly alters the training models. In this work, a novel method for reducing the effect of label noise is introduced. The rules are created from expert k...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Quoc Nguyen, Tomoaki Shikina, Daichi Teruya, Seiji Hotta, Huy-Dung Han, Hironori Nakajo
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	prior knowledge noise datasets label noise weighting training data Technology T Engineering (General). Civil engineering (General) TA1-2040 Biology (General) QH301-705.5 Physics QC1-999 Chemistry QD1-999
Acceso en línea:	https://doaj.org/article/10b06feac2404b61b1fc2aab9d800a26
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:10b06feac2404b61b1fc2aab9d800a26
record_format	dspace
spelling	oai:doaj.org-article:10b06feac2404b61b1fc2aab9d800a262021-11-25T16:43:13ZLeveraging Expert Knowledge for Label Noise Mitigation in Machine Learning10.3390/app1122110402076-3417https://doaj.org/article/10b06feac2404b61b1fc2aab9d800a262021-11-01T00:00:00Zhttps://www.mdpi.com/2076-3417/11/22/11040https://doaj.org/toc/2076-3417In training-based Machine Learning applications, the training data are frequently labeled by non-experts and expose substantial label noise which greatly alters the training models. In this work, a novel method for reducing the effect of label noise is introduced. The rules are created from expert knowledge to identify the incorrect non-expert training data. Using the gradient descent algorithm, the violating data samples are weighted less to mitigate their effects during model training. The proposed method is applied to the image classification problem using Manga109 and CIFAR-10 dataset. The experiments show that when the noise level is up to 50% our proposed method significantly increases the accuracy of the model compared to conventional learning methods.Quoc NguyenTomoaki ShikinaDaichi TeruyaSeiji HottaHuy-Dung HanHironori NakajoMDPI AGarticleprior knowledgenoise datasetslabel noiseweighting training dataTechnologyTEngineering (General). Civil engineering (General)TA1-2040Biology (General)QH301-705.5PhysicsQC1-999ChemistryQD1-999ENApplied Sciences, Vol 11, Iss 11040, p 11040 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	prior knowledge noise datasets label noise weighting training data Technology T Engineering (General). Civil engineering (General) TA1-2040 Biology (General) QH301-705.5 Physics QC1-999 Chemistry QD1-999
spellingShingle	prior knowledge noise datasets label noise weighting training data Technology T Engineering (General). Civil engineering (General) TA1-2040 Biology (General) QH301-705.5 Physics QC1-999 Chemistry QD1-999 Quoc Nguyen Tomoaki Shikina Daichi Teruya Seiji Hotta Huy-Dung Han Hironori Nakajo Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
description	In training-based Machine Learning applications, the training data are frequently labeled by non-experts and expose substantial label noise which greatly alters the training models. In this work, a novel method for reducing the effect of label noise is introduced. The rules are created from expert knowledge to identify the incorrect non-expert training data. Using the gradient descent algorithm, the violating data samples are weighted less to mitigate their effects during model training. The proposed method is applied to the image classification problem using Manga109 and CIFAR-10 dataset. The experiments show that when the noise level is up to 50% our proposed method significantly increases the accuracy of the model compared to conventional learning methods.
format	article
author	Quoc Nguyen Tomoaki Shikina Daichi Teruya Seiji Hotta Huy-Dung Han Hironori Nakajo
author_facet	Quoc Nguyen Tomoaki Shikina Daichi Teruya Seiji Hotta Huy-Dung Han Hironori Nakajo
author_sort	Quoc Nguyen
title	Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
title_short	Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
title_full	Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
title_fullStr	Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
title_full_unstemmed	Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning
title_sort	leveraging expert knowledge for label noise mitigation in machine learning
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/10b06feac2404b61b1fc2aab9d800a26
work_keys_str_mv	AT quocnguyen leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning AT tomoakishikina leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning AT daichiteruya leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning AT seijihotta leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning AT huydunghan leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning AT hironorinakajo leveragingexpertknowledgeforlabelnoisemitigationinmachinelearning
_version_	1718413000418787328

Leveraging Expert Knowledge for Label Noise Mitigation in Machine Learning

Ejemplares similares