Transfer learning for galaxy morphology from one survey to another

Transfer learning for galaxy morphology from one survey to another
复制标题

星系形态从一项调查到另一项调查的迁移学习

DOI:
10.1093/mnras/sty3497
复制
发表时间:
2018
影响因子:
4.8
通讯作者:
Brooks, D
Brooks, D
中科院分区:
物理与天体物理2区
文献类型:
--
作者:
Domínguez Sánchez, H;Huertas-Company, M;Bernardi, M;Kaviraj, S;Fischer, J L;Abbott, T M;Abdalla, F B;Annis, J;Avila, S;Brooks, D

文献摘要

参考文献

被引文献

相似文献

事实证明,用于星系形态分类的深度学习(DL)算法非常成功,模仿(甚至改进)了视觉分类。然而,这些算法依赖于标记星系的大量训练样本(通常有数千个)。在未来的大数据调查中使用深度学习分类的一个关键问题是,从现有调查中获得的知识有多少可以导出到新的数据集,即机器学习的特征是否对不同的数据有意义。我们测试的DL模型的性能,训练与斯隆数字巡天(SDSS)的数据,暗能量调查(DES)使用的图像为样本的15000个星系与类似的红移分布SDSS。将模型直接应用于DES数据可提供合理的总体准确度(约90%),但完整性和纯度值较小。一个快速的域适应步骤,包括一个小的DES样本的星系(500-300)的进一步培训,是足够的,以获得一个准确度> 95%,并显着改善的完整性和纯度值。这表明,一旦使用特定的数据集进行训练,机器就可以快速适应新的仪器特征(例如PSF,视宁度,深度),从而将形态分类所需的训练样本减少近一个数量级。在本研究中没有考虑红移演化效应或显著的深度差异。
Deep learning (DL) algorithms for morphological classification of galaxies have proven very successful, mimicking (or even improving) visual classifications. However, these algorithms rely on large training samples of labelled galaxies (typically thousands of them). A key question for using DL classifications in future Big Data surveys is how much of the knowledge acquired from an existing survey can be exported to a new data set, i.e. if the features learned by the machines are meaningful for different data. We test the performance of DL models, trained with Sloan Digital Sky Survey (SDSS) data, on Dark Energy Survey (DES) using images for a sample of ∼5000 galaxies with a similar redshift distribution to SDSS. Applying the models directly to DES data provides a reasonable global accuracy (∼90 per cent), but small completeness and purity values. A fast domain adaptation step, consisting of a further training with a small DES sample of galaxies (∼500–300), is enough for obtaining an accuracy >95 per cent and a significant improvement in the completeness and purity values. This demonstrates that, once trained with a particular data set, machines can quickly adapt to new instrument characteristics (e.g. PSF, seeing, depth), reducing by almost one order of magnitude the necessary training sample for morphological classification. Redshift evolution effects or significant depth differences are not taken into account in this study.
DOI: 10.1088/0067-0049/199/2/25
发表时间: 2011-06
期刊: The Astrophysical Journal Supplement Series
影响因子: --
作者:
M. Postman;D. Coe;N. Benı́tez;L. Bradley;T. Broadhurst;M. Donahue;H. Ford;O. Graur;G. Graves-G.-G
通讯作者: M. Postman;D. Coe;N. Benı́tez;L. Bradley;T. Broadhurst;M. Donahue;H. Ford;O. Graur;G. Graves-G.-G
使用迁移学习来检测星系合并
DOI: 10.1093/mnras/sty1398
发表时间: 2018
期刊: ArXiv
影响因子: --
作者:
Sandro Ackermann;K. Schawinski;Ce Zhang;Anna K. Weigel;M. D. Turp
通讯作者: M. D. Turp
DOI: 10.1088/0067-0049/221/1/8
发表时间: 2015-11-01
影响因子: 8.7
作者:
Huertas-Company, M.;Gravet, R.;Mcintosh, D. H.
通讯作者: Mcintosh, D. H.
加州理工学院/加州大学圣地亚哥分校(美国)
DOI: --
发表时间: --
期刊:
影响因子: --
作者:
通讯作者: --
“与组方案相关的角色产品和平衡集”(预印本)。
DOI: --
发表时间: --
期刊:
影响因子: --
作者:
通讯作者: --