Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web

In our previous works, we proposed a machine learning framework named USmorph for efficiently classifying galaxy morphology. In this study, we propose a self-supervised method called contrastive learning to upgrade the unsupervised machine learning (UML) part of the USmorph framework, aiming to impr...

Full description

Saved in:
Bibliographic Details
Main Authors: Shiwei Zhu, Guanwen Fang, Chichun Zhou, Jie Song, Zesen Lin, Yao Dai, Xu Kong
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:The Astrophysical Journal Supplement Series
Subjects:
Online Access:https://doi.org/10.3847/1538-4365/add0b8
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850108484367941632
author Shiwei Zhu
Guanwen Fang
Chichun Zhou
Jie Song
Zesen Lin
Yao Dai
Xu Kong
author_facet Shiwei Zhu
Guanwen Fang
Chichun Zhou
Jie Song
Zesen Lin
Yao Dai
Xu Kong
author_sort Shiwei Zhu
collection DOAJ
description In our previous works, we proposed a machine learning framework named USmorph for efficiently classifying galaxy morphology. In this study, we propose a self-supervised method called contrastive learning to upgrade the unsupervised machine learning (UML) part of the USmorph framework, aiming to improve the efficiency of feature extraction in this step. The upgraded UML method primarily consists of the following three aspects. (1) We employ a convolutional autoencoder to denoise galaxy images and adaptive polar coordinate transformation to enhance the model’s rotational invariance. (2) A pretrained dual-encoder convolutional neural network based on ConvNeXt and a vision transformer is used to encode the image data, while contrastive learning is then applied to reduce the dimension of the features. (3) We adopt a bagging-based clustering model to cluster galaxies with similar features into distinct groups. By carefully dividing the redshift bins, we apply this model to the rest-frame optical images of galaxies in the COSMOS-Web field within the redshift range of 0.5 <  z  < 6.0. Compared to the previous algorithm, the improved UML method successfully classifies 73% of galaxies. Using the GoogLeNet algorithm, we classify the morphology of the remaining 27% of galaxies. To validate the reliability of our updated algorithm, we compared our classification results with other galaxy morphological parameters and found a good consistency with galaxy evolution. Benefiting from its higher efficiency, this updated algorithm is well suited for application in future China Space Station Telescope missions.
format Article
id doaj-art-72af8f1eb7b44b11ba8cf59def4b2f2f
institution OA Journals
issn 0067-0049
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series The Astrophysical Journal Supplement Series
spelling doaj-art-72af8f1eb7b44b11ba8cf59def4b2f2f2025-08-20T02:38:21ZengIOP PublishingThe Astrophysical Journal Supplement Series0067-00492025-01-0127823910.3847/1538-4365/add0b8Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-WebShiwei Zhu0https://orcid.org/0009-0004-0966-6439Guanwen Fang1https://orcid.org/0000-0001-9694-2171Chichun Zhou2https://orcid.org/0000-0002-5133-2668Jie Song3https://orcid.org/0000-0002-0846-7591Zesen Lin4https://orcid.org/0000-0001-8078-3428Yao Dai5https://orcid.org/0000-0002-4638-0235Xu Kong6https://orcid.org/0000-0002-7660-2273School of Mathematics and Physics, Anqing Normal University , Anqing 246011, People’s Republic of China ; wen@mail.ustc.edu.cn; Institute of Astronomy and Astrophysics, Anqing Normal University , Anqing 246133, People’s Republic of ChinaSchool of Mathematics and Physics, Anqing Normal University , Anqing 246011, People’s Republic of China ; wen@mail.ustc.edu.cn; Institute of Astronomy and Astrophysics, Anqing Normal University , Anqing 246133, People’s Republic of ChinaSchool of Engineering, Dali University , Dali 671003, People’s Republic of ChinaDepartment of Astronomy, University of Science and Technology of China , Hefei 230026, People’s Republic of China ; xkong@ustc.edu.cn; School of Astronomy and Space Science, University of Science and Technology of China , Hefei 230026, People’s Republic of China; Institute of Deep Space Sciences , Deep Space Exploration Laboratory, Hefei 230026, People’s Republic of ChinaDepartment of Physics, The Chinese University of Hong Kong , Shatin, N.T., Hong Kong S.A.R., People’s Republic of ChinaShanghai Astronomical Observatory, Chinese Academy of Sciences , 80 Nandan Road, Shanghai 200030, People’s Republic of China; School of Astronomy and Space Science, University of Chinese Academy of Sciences , No. 19A Yuquan Road, Beijing 100049, People’s Republic of ChinaDepartment of Astronomy, University of Science and Technology of China , Hefei 230026, People’s Republic of China ; xkong@ustc.edu.cn; School of Astronomy and Space Science, University of Science and Technology of China , Hefei 230026, People’s Republic of China; Institute of Deep Space Sciences , Deep Space Exploration Laboratory, Hefei 230026, People’s Republic of ChinaIn our previous works, we proposed a machine learning framework named USmorph for efficiently classifying galaxy morphology. In this study, we propose a self-supervised method called contrastive learning to upgrade the unsupervised machine learning (UML) part of the USmorph framework, aiming to improve the efficiency of feature extraction in this step. The upgraded UML method primarily consists of the following three aspects. (1) We employ a convolutional autoencoder to denoise galaxy images and adaptive polar coordinate transformation to enhance the model’s rotational invariance. (2) A pretrained dual-encoder convolutional neural network based on ConvNeXt and a vision transformer is used to encode the image data, while contrastive learning is then applied to reduce the dimension of the features. (3) We adopt a bagging-based clustering model to cluster galaxies with similar features into distinct groups. By carefully dividing the redshift bins, we apply this model to the rest-frame optical images of galaxies in the COSMOS-Web field within the redshift range of 0.5 <  z  < 6.0. Compared to the previous algorithm, the improved UML method successfully classifies 73% of galaxies. Using the GoogLeNet algorithm, we classify the morphology of the remaining 27% of galaxies. To validate the reliability of our updated algorithm, we compared our classification results with other galaxy morphological parameters and found a good consistency with galaxy evolution. Benefiting from its higher efficiency, this updated algorithm is well suited for application in future China Space Station Telescope missions.https://doi.org/10.3847/1538-4365/add0b8Galaxy structureAstrostatistics techniquesAstronomy data analysis
spellingShingle Shiwei Zhu
Guanwen Fang
Chichun Zhou
Jie Song
Zesen Lin
Yao Dai
Xu Kong
Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
The Astrophysical Journal Supplement Series
Galaxy structure
Astrostatistics techniques
Astronomy data analysis
title Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
title_full Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
title_fullStr Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
title_full_unstemmed Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
title_short Dual-coding Contrastive Learning Based on the ConvNeXt and ViT Models for Morphological Classification of Galaxies in COSMOS-Web
title_sort dual coding contrastive learning based on the convnext and vit models for morphological classification of galaxies in cosmos web
topic Galaxy structure
Astrostatistics techniques
Astronomy data analysis
url https://doi.org/10.3847/1538-4365/add0b8
work_keys_str_mv AT shiweizhu dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT guanwenfang dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT chichunzhou dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT jiesong dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT zesenlin dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT yaodai dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb
AT xukong dualcodingcontrastivelearningbasedontheconvnextandvitmodelsformorphologicalclassificationofgalaxiesincosmosweb