Active Learning of Pattern Classification Based on PEDCC-Loss

Deep learning classifiers require a large number of labeled samples to train the model. Active learning reduces the dependence of classification model on labeled samples by gradually selecting high-quality samples for iterative training. In this article, an active learning method for pattern classif...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Qiuyu Zhu, Jianbing Luan, Tiantian Li, Xuewen Zu
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Active learning predefined class centroids PEDCC-loss pattern classification Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/79ec406f64c141e78db6def879c1bb4b
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:79ec406f64c141e78db6def879c1bb4b
record_format	dspace
spelling	oai:doaj.org-article:79ec406f64c141e78db6def879c1bb4b2021-11-18T00:10:19ZActive Learning of Pattern Classification Based on PEDCC-Loss2169-353610.1109/ACCESS.2021.3123095https://doaj.org/article/79ec406f64c141e78db6def879c1bb4b2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9585442/https://doaj.org/toc/2169-3536Deep learning classifiers require a large number of labeled samples to train the model. Active learning reduces the dependence of classification model on labeled samples by gradually selecting high-quality samples for iterative training. In this article, an active learning method for pattern classification is proposed, which can use fewer labeled samples to achieve higher classification accuracy. The algorithm uses PEDCC-Loss to train the network model according to the predefined evenly-distributed class centroids (PEDCC) of the classification task, maximizing the distance between classes. Then, the maximum value of the cosine distance is selected to measure the sample uncertainty according to the PEDCC output characteristics. In the active learning algorithm iteration, samples with high uncertainty are selected for manual labeling, and samples with low uncertainty are given pseudo labels. At the same time, manual label samples and pseudo label samples are used to train the classification network to continuously optimize the classification boundary. Compared with CRL, LLAL and other methods, the experimental results on CIFAR100, CIFAR10, SVHN, and Animals-10 data sets show the effectiveness of this algorithm.Qiuyu ZhuJianbing LuanTiantian LiXuewen ZuIEEEarticleActive learningpredefined class centroidsPEDCC-losspattern classificationElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 147626-147633 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Active learning predefined class centroids PEDCC-loss pattern classification Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Active learning predefined class centroids PEDCC-loss pattern classification Electrical engineering. Electronics. Nuclear engineering TK1-9971 Qiuyu Zhu Jianbing Luan Tiantian Li Xuewen Zu Active Learning of Pattern Classification Based on PEDCC-Loss
description	Deep learning classifiers require a large number of labeled samples to train the model. Active learning reduces the dependence of classification model on labeled samples by gradually selecting high-quality samples for iterative training. In this article, an active learning method for pattern classification is proposed, which can use fewer labeled samples to achieve higher classification accuracy. The algorithm uses PEDCC-Loss to train the network model according to the predefined evenly-distributed class centroids (PEDCC) of the classification task, maximizing the distance between classes. Then, the maximum value of the cosine distance is selected to measure the sample uncertainty according to the PEDCC output characteristics. In the active learning algorithm iteration, samples with high uncertainty are selected for manual labeling, and samples with low uncertainty are given pseudo labels. At the same time, manual label samples and pseudo label samples are used to train the classification network to continuously optimize the classification boundary. Compared with CRL, LLAL and other methods, the experimental results on CIFAR100, CIFAR10, SVHN, and Animals-10 data sets show the effectiveness of this algorithm.
format	article
author	Qiuyu Zhu Jianbing Luan Tiantian Li Xuewen Zu
author_facet	Qiuyu Zhu Jianbing Luan Tiantian Li Xuewen Zu
author_sort	Qiuyu Zhu
title	Active Learning of Pattern Classification Based on PEDCC-Loss
title_short	Active Learning of Pattern Classification Based on PEDCC-Loss
title_full	Active Learning of Pattern Classification Based on PEDCC-Loss
title_fullStr	Active Learning of Pattern Classification Based on PEDCC-Loss
title_full_unstemmed	Active Learning of Pattern Classification Based on PEDCC-Loss
title_sort	active learning of pattern classification based on pedcc-loss
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/79ec406f64c141e78db6def879c1bb4b
work_keys_str_mv	AT qiuyuzhu activelearningofpatternclassificationbasedonpedccloss AT jianbingluan activelearningofpatternclassificationbasedonpedccloss AT tiantianli activelearningofpatternclassificationbasedonpedccloss AT xuewenzu activelearningofpatternclassificationbasedonpedccloss
_version_	1718425231684534272

Active Learning of Pattern Classification Based on PEDCC-Loss

Ejemplares similares