A novel semi-supervised learning method based on fast search and density peaks

Fei Gao, Teng Huang, Jinping Sun, Amir Hussain, Erfu Yang, Huiyu Zhou

Research output: Contribution to journalArticle

Abstract

Radar image recognition is a hotspot in the field of remote sensing. Under the condition of sufficiently labeled samples, recognition algorithms can achieve good classification results. However, labeled samples are scarce and costly to obtain. Our major interest in this paper is how to use these unlabeled samples to improve the performance of a recognition algorithm in the case of limited labeled samples. This is a semi-supervised learning problem. However, unlike the existing semi-supervised learning methods, we do not use unlabeled samples directly and, instead, look for safe and reliable unlabeled samples before using them. In this paper, two new semi-supervised learning methods are proposed: a semi-supervised learning method based on fast search and density peaks (S 2 DP) and an iterative S 2 DP method (IS 2 DP). When the labeled samples satisfy a certain requirement, S2DP uses fast search and a density peak clustering method to detect reliable unlabeled samples based on the weighted kernel Fisher discriminant analysis (WKFDA). Then, a labeling method based on clustering information (LCI) is designed to label the unlabeled samples. When the labeled samples are insufficient, IS2DP is used to iteratively search for reliable unlabeled samples for semi-supervision. Then, these samples are added to the labeled samples to improve the recognition performance of S2DP. In the experiments, real radar images are used to verify the performance of our proposed algorithm in dealing with the scarcity of the labeled samples. In addition, our algorithm is compared against several semi-supervised deep learning methods with similar structures. Experimental results demonstrate that the proposed algorithm has better stability than these methods.

LanguageEnglish
Article number6876173
Pages1-23
Number of pages23
JournalComplexity
Volume2019
DOIs
Publication statusPublished - 3 Feb 2019

Fingerprint

Supervised learning
Radar
Image recognition
Discriminant analysis
Labeling
Labels
Remote sensing
Semi-supervised learning
Learning methods
Experiments

Keywords

  • remote sensing
  • radar image recognition
  • recognition algorithm

Cite this

Gao, Fei ; Huang, Teng ; Sun, Jinping ; Hussain, Amir ; Yang, Erfu ; Zhou, Huiyu. / A novel semi-supervised learning method based on fast search and density peaks. In: Complexity. 2019 ; Vol. 2019. pp. 1-23.
@article{3b132a03a01f4747b4032aaa847b0a38,
title = "A novel semi-supervised learning method based on fast search and density peaks",
abstract = "Radar image recognition is a hotspot in the field of remote sensing. Under the condition of sufficiently labeled samples, recognition algorithms can achieve good classification results. However, labeled samples are scarce and costly to obtain. Our major interest in this paper is how to use these unlabeled samples to improve the performance of a recognition algorithm in the case of limited labeled samples. This is a semi-supervised learning problem. However, unlike the existing semi-supervised learning methods, we do not use unlabeled samples directly and, instead, look for safe and reliable unlabeled samples before using them. In this paper, two new semi-supervised learning methods are proposed: a semi-supervised learning method based on fast search and density peaks (S 2 DP) and an iterative S 2 DP method (IS 2 DP). When the labeled samples satisfy a certain requirement, S2DP uses fast search and a density peak clustering method to detect reliable unlabeled samples based on the weighted kernel Fisher discriminant analysis (WKFDA). Then, a labeling method based on clustering information (LCI) is designed to label the unlabeled samples. When the labeled samples are insufficient, IS2DP is used to iteratively search for reliable unlabeled samples for semi-supervision. Then, these samples are added to the labeled samples to improve the recognition performance of S2DP. In the experiments, real radar images are used to verify the performance of our proposed algorithm in dealing with the scarcity of the labeled samples. In addition, our algorithm is compared against several semi-supervised deep learning methods with similar structures. Experimental results demonstrate that the proposed algorithm has better stability than these methods.",
keywords = "remote sensing, radar image recognition, recognition algorithm",
author = "Fei Gao and Teng Huang and Jinping Sun and Amir Hussain and Erfu Yang and Huiyu Zhou",
year = "2019",
month = "2",
day = "3",
doi = "10.1155/2019/6876173",
language = "English",
volume = "2019",
pages = "1--23",
journal = "Emergence",
issn = "1521-3250",

}

A novel semi-supervised learning method based on fast search and density peaks. / Gao, Fei; Huang, Teng; Sun, Jinping; Hussain, Amir; Yang, Erfu; Zhou, Huiyu.

In: Complexity, Vol. 2019, 6876173, 03.02.2019, p. 1-23.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A novel semi-supervised learning method based on fast search and density peaks

AU - Gao, Fei

AU - Huang, Teng

AU - Sun, Jinping

AU - Hussain, Amir

AU - Yang, Erfu

AU - Zhou, Huiyu

PY - 2019/2/3

Y1 - 2019/2/3

N2 - Radar image recognition is a hotspot in the field of remote sensing. Under the condition of sufficiently labeled samples, recognition algorithms can achieve good classification results. However, labeled samples are scarce and costly to obtain. Our major interest in this paper is how to use these unlabeled samples to improve the performance of a recognition algorithm in the case of limited labeled samples. This is a semi-supervised learning problem. However, unlike the existing semi-supervised learning methods, we do not use unlabeled samples directly and, instead, look for safe and reliable unlabeled samples before using them. In this paper, two new semi-supervised learning methods are proposed: a semi-supervised learning method based on fast search and density peaks (S 2 DP) and an iterative S 2 DP method (IS 2 DP). When the labeled samples satisfy a certain requirement, S2DP uses fast search and a density peak clustering method to detect reliable unlabeled samples based on the weighted kernel Fisher discriminant analysis (WKFDA). Then, a labeling method based on clustering information (LCI) is designed to label the unlabeled samples. When the labeled samples are insufficient, IS2DP is used to iteratively search for reliable unlabeled samples for semi-supervision. Then, these samples are added to the labeled samples to improve the recognition performance of S2DP. In the experiments, real radar images are used to verify the performance of our proposed algorithm in dealing with the scarcity of the labeled samples. In addition, our algorithm is compared against several semi-supervised deep learning methods with similar structures. Experimental results demonstrate that the proposed algorithm has better stability than these methods.

AB - Radar image recognition is a hotspot in the field of remote sensing. Under the condition of sufficiently labeled samples, recognition algorithms can achieve good classification results. However, labeled samples are scarce and costly to obtain. Our major interest in this paper is how to use these unlabeled samples to improve the performance of a recognition algorithm in the case of limited labeled samples. This is a semi-supervised learning problem. However, unlike the existing semi-supervised learning methods, we do not use unlabeled samples directly and, instead, look for safe and reliable unlabeled samples before using them. In this paper, two new semi-supervised learning methods are proposed: a semi-supervised learning method based on fast search and density peaks (S 2 DP) and an iterative S 2 DP method (IS 2 DP). When the labeled samples satisfy a certain requirement, S2DP uses fast search and a density peak clustering method to detect reliable unlabeled samples based on the weighted kernel Fisher discriminant analysis (WKFDA). Then, a labeling method based on clustering information (LCI) is designed to label the unlabeled samples. When the labeled samples are insufficient, IS2DP is used to iteratively search for reliable unlabeled samples for semi-supervision. Then, these samples are added to the labeled samples to improve the recognition performance of S2DP. In the experiments, real radar images are used to verify the performance of our proposed algorithm in dealing with the scarcity of the labeled samples. In addition, our algorithm is compared against several semi-supervised deep learning methods with similar structures. Experimental results demonstrate that the proposed algorithm has better stability than these methods.

KW - remote sensing

KW - radar image recognition

KW - recognition algorithm

UR - http://www.scopus.com/inward/record.url?scp=85062348489&partnerID=8YFLogxK

U2 - 10.1155/2019/6876173

DO - 10.1155/2019/6876173

M3 - Article

VL - 2019

SP - 1

EP - 23

JO - Emergence

T2 - Emergence

JF - Emergence

SN - 1521-3250

M1 - 6876173

ER -