Transfer learning for galaxy morphology from one survey to another

H. Doḿinguez Sanchez; M. Huertas-Company; M. Bernardi; S. Kaviraj; J. L. Fischer; T. M.C. Abbott; F. B. Abdalla; J. Annis; S. Avila; D. Brooks; E. Buckley-Geer; A. Carnero Rosell; M. Carrasco Kind; J. Carretero; C. E. Cunha; C. B. D'Andrea; L. N. Da Costa; C. Davis; J. De Vicente; P. Doel; A. E. Evrard; P. Fosalba; J. Frieman; J. Garćia-Bellido; E. Gaztanaga; D. W. Gerdes; D. Gruen; R. A. Gruendl; J. Gschwend; G. Gutierrez; W. G. Hartley; D. L. Hollowood; K. Honscheid; B. Hoyle; D. J. James; K. Kuehn; N. Kuropatkin; O. Lahav; M. A.G. Maia; M. March; P. Melchior; F. Menanteau; R. Miquel; B. Nord; A. A. Plazas; E. Sanchez; V. Scarpine; R. Schindler; M. Schubnell; M. Smith; R. C. Smith; M. Soares-Santos; F. Sobreira; E. Suchyta; M. E.C. Swanson; G. Tarle; D. Thomas; A. R. Walker; J. Zuntz

doi:10.1093/mnras/sty3497

Transfer learning for galaxy morphology from one survey to another

H. Doḿinguez Sanchez, M. Huertas-Company, M. Bernardi, S. Kaviraj, J. L. Fischer, T. M.C. Abbott, F. B. Abdalla, J. Annis, S. Avila, D. Brooks, E. Buckley-Geer, A. Carnero Rosell, M. Carrasco Kind, J. Carretero, C. E. Cunha, C. B. D'Andrea, L. N. Da Costa, C. Davis, J. De Vicente, P. DoelA. E. Evrard, P. Fosalba, J. Frieman, J. Garćia-Bellido, E. Gaztanaga, D. W. Gerdes, D. Gruen, R. A. Gruendl, J. Gschwend, G. Gutierrez, W. G. Hartley, D. L. Hollowood, K. Honscheid, B. Hoyle, D. J. James, K. Kuehn, N. Kuropatkin, O. Lahav, M. A.G. Maia, M. March, P. Melchior, F. Menanteau, R. Miquel, B. Nord, A. A. Plazas, E. Sanchez, V. Scarpine, R. Schindler, M. Schubnell, M. Smith, R. C. Smith, M. Soares-Santos, F. Sobreira, E. Suchyta, M. E.C. Swanson, G. Tarle, D. Thomas, A. R. Walker, J. Zuntz

Workflow Systems

Research output: Contribution to journal › Article › peer-review

28 Scopus citations

Abstract

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.

Original language	English
Pages (from-to)	93-100
Number of pages	8
Journal	Monthly Notices of the Royal Astronomical Society
Volume	484
Issue number	1
DOIs	https://doi.org/10.1093/mnras/sty3497
State	Published - Mar 21 2019

Bibliographical note

Publisher Copyright:
© 2018 The Author(s).

Funding

The authors would like to thank the anonymous referee for the useful comments which helped to improve the content of the paper. This work was supported in part by NSF AST-1816330. Funding for the DES Projects has been provided by the DOE and NSF(USA), MEC/MICINN/MINECO(Spain), STFC(UK), HEFCE(UK). NCSA(UIUC), KICP(U. Chicago), CCAPP(Ohio State), MIFPA(Texas A&M), CNPQ, FAPERJ, FINEP (Brazil), DFG(Germany), and the Collaborating Institutions in the Dark Energy Survey. The Collaborating Institutions are Argonne Lab, UC Santa Cruz, University of Cambridge, CIEMAT-Madrid, University of Chicago, University College London, DES-Brazil Consortium, University of Edinburgh, ETH Zürich, Fermilab, University of Illinois, ICE (IEEC-CSIC), IFAE Barcelona, Lawrence Berkeley Lab, LMU München and the associated Excellence Cluster Universe, University of Michigan, NOAO, University of Nottingham, Ohio State University, University of Pennsylvania, University of Portsmouth, SLAC National Lab, Stanford University, University of Sussex, Texas A&M University, and the OzDES Membership Consortium. Based in part on observations at Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. The DES Data Management System is supported by the NSF under Grant Numbers AST-1138766 and AST-1536171. The DES participants from Spanish institutions are partially supported by MINECO under grants AYA2015-71825, ESP2015-66861, FPA2015-68048, SEV-2016-0588, SEV-2016-0597, and MDM-2015-0509, some of which include ERDF funds from the European Union. IFAE is partially funded by the CERCA program of the Generalitat de Catalunya. Research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Program (FP7/2007-2013) including ERC grant agreements 240672, 291329, and 306478. We acknowledge support from the Australian Research Council Centre of Excellence for All-sky Astrophysics (CAAS-TRO), through project number CE110001020, and the Brazilian Instituto Nacional de Ciência e Tecnologia (INCT) e-Universe (CNPq grant 465376/2014-2). This manuscript has been authored by Fermi Research Alliance, LLC under Contract No. DE-AC02-07CH11359 with the U.S. Department of Energy, Office of Science, Office of High Energy Physics. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a nonexclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.

Funders	Funder number
Brazilian Instituto Nacional de Ciência e Tecnologia
CAAS-TRO	CE110001020
Fermi Research Alliance, LLC	DE-AC02-07CH11359
INCT
National Science Foundation	SEV-2016-0588, SEV-2016-0597, ESP2015-66861, 1138766, AST-1138766, MDM-2015-0509, 1816330, FPA2015-68048, AST-1536171, AYA2015-71825, AST-1816330
U.S. Department of Energy
Office of Science
High Energy Physics
Seventh Framework Programme	240672, 306478, 291329
Engineering Research Centers
Science and Technology Facilities Council	ST/N000668/1
European Commission
European Research Council
Australian Research Council
Generalitat de Catalunya
Ministerio de Economía y Competitividad
Conselho Nacional de Desenvolvimento Científico e Tecnológico	465376/2014-2
Ministerio de Ciencia e Innovación
European Regional Development Fund

Keywords

Galaxies: structure
Methods: observational
Methods: photometric
Surveys

Access to Document

10.1093/mnras/sty3497

Cite this

Doḿinguez Sanchez, H., Huertas-Company, M., Bernardi, M., Kaviraj, S., Fischer, J. L., Abbott, T. M. C., Abdalla, F. B., Annis, J., Avila, S., Brooks, D., Buckley-Geer, E., Carnero Rosell, A., Carrasco Kind, M., Carretero, J., Cunha, C. E., D'Andrea, C. B., Da Costa, L. N., Davis, C., De Vicente, J., ... Zuntz, J. (2019). Transfer learning for galaxy morphology from one survey to another. Monthly Notices of the Royal Astronomical Society, 484(1), 93-100. https://doi.org/10.1093/mnras/sty3497

@article{9049881f9d204b05987d1a81341a5606,

title = "Transfer learning for galaxy morphology from one survey to another",

abstract = "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.",

keywords = "Galaxies: structure, Methods: observational, Methods: photometric, Surveys",

author = "{Doḿinguez Sanchez}, H. and M. Huertas-Company and M. Bernardi and S. Kaviraj and Fischer, {J. L.} and Abbott, {T. M.C.} and Abdalla, {F. B.} and J. Annis and S. Avila and D. Brooks and E. Buckley-Geer and {Carnero Rosell}, A. and {Carrasco Kind}, M. and J. Carretero and Cunha, {C. E.} and D'Andrea, {C. B.} and {Da Costa}, {L. N.} and C. Davis and {De Vicente}, J. and P. Doel and Evrard, {A. E.} and P. Fosalba and J. Frieman and J. Gar{\'c}ia-Bellido and E. Gaztanaga and Gerdes, {D. W.} and D. Gruen and Gruendl, {R. A.} and J. Gschwend and G. Gutierrez and Hartley, {W. G.} and Hollowood, {D. L.} and K. Honscheid and B. Hoyle and James, {D. J.} and K. Kuehn and N. Kuropatkin and O. Lahav and Maia, {M. A.G.} and M. March and P. Melchior and F. Menanteau and R. Miquel and B. Nord and Plazas, {A. A.} and E. Sanchez and V. Scarpine and R. Schindler and M. Schubnell and M. Smith and Smith, {R. C.} and M. Soares-Santos and F. Sobreira and E. Suchyta and Swanson, {M. E.C.} and G. Tarle and D. Thomas and Walker, {A. R.} and J. Zuntz",

note = "Publisher Copyright: {\textcopyright} 2018 The Author(s).",

year = "2019",

month = mar,

day = "21",

doi = "10.1093/mnras/sty3497",

language = "English",

volume = "484",

pages = "93--100",

journal = "Monthly Notices of the Royal Astronomical Society",

issn = "0035-8711",

publisher = "Oxford University Press",

number = "1",

}

Doḿinguez Sanchez, H, Huertas-Company, M, Bernardi, M, Kaviraj, S, Fischer, JL, Abbott, TMC, Abdalla, FB, Annis, J, Avila, S, Brooks, D, Buckley-Geer, E, Carnero Rosell, A, Carrasco Kind, M, Carretero, J, Cunha, CE, D'Andrea, CB, Da Costa, LN, Davis, C, De Vicente, J, Doel, P, Evrard, AE, Fosalba, P, Frieman, J, Garćia-Bellido, J, Gaztanaga, E, Gerdes, DW, Gruen, D, Gruendl, RA, Gschwend, J, Gutierrez, G, Hartley, WG, Hollowood, DL, Honscheid, K, Hoyle, B, James, DJ, Kuehn, K, Kuropatkin, N, Lahav, O, Maia, MAG, March, M, Melchior, P, Menanteau, F, Miquel, R, Nord, B, Plazas, AA, Sanchez, E, Scarpine, V, Schindler, R, Schubnell, M, Smith, M, Smith, RC, Soares-Santos, M, Sobreira, F, Suchyta, E, Swanson, MEC, Tarle, G, Thomas, D, Walker, AR & Zuntz, J 2019, 'Transfer learning for galaxy morphology from one survey to another', Monthly Notices of the Royal Astronomical Society, vol. 484, no. 1, pp. 93-100. https://doi.org/10.1093/mnras/sty3497

TY - JOUR

T1 - Transfer learning for galaxy morphology from one survey to another

AU - Doḿinguez Sanchez, H.

AU - Huertas-Company, M.

AU - Bernardi, M.

AU - Kaviraj, S.

AU - Fischer, J. L.

AU - Abbott, T. M.C.

AU - Abdalla, F. B.

AU - Annis, J.

AU - Avila, S.

AU - Brooks, D.

AU - Buckley-Geer, E.

AU - Carnero Rosell, A.

AU - Carrasco Kind, M.

AU - Carretero, J.

AU - Cunha, C. E.

AU - D'Andrea, C. B.

AU - Da Costa, L. N.

AU - Davis, C.

AU - De Vicente, J.

AU - Doel, P.

AU - Evrard, A. E.

AU - Fosalba, P.

AU - Frieman, J.

AU - Garćia-Bellido, J.

AU - Gaztanaga, E.

AU - Gerdes, D. W.

AU - Gruen, D.

AU - Gruendl, R. A.

AU - Gschwend, J.

AU - Gutierrez, G.

AU - Hartley, W. G.

AU - Hollowood, D. L.

AU - Honscheid, K.

AU - Hoyle, B.

AU - James, D. J.

AU - Kuehn, K.

AU - Kuropatkin, N.

AU - Lahav, O.

AU - Maia, M. A.G.

AU - March, M.

AU - Melchior, P.

AU - Menanteau, F.

AU - Miquel, R.

AU - Nord, B.

AU - Plazas, A. A.

AU - Sanchez, E.

AU - Scarpine, V.

AU - Schindler, R.

AU - Schubnell, M.

AU - Smith, M.

AU - Smith, R. C.

AU - Soares-Santos, M.

AU - Sobreira, F.

AU - Suchyta, E.

AU - Swanson, M. E.C.

AU - Tarle, G.

AU - Thomas, D.

AU - Walker, A. R.

AU - Zuntz, J.

PY - 2019/3/21

Y1 - 2019/3/21

N2 - 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.

AB - 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.

KW - Galaxies: structure

KW - Methods: observational

KW - Methods: photometric

KW - Surveys

UR - http://www.scopus.com/inward/record.url?scp=85062264285&partnerID=8YFLogxK

U2 - 10.1093/mnras/sty3497

DO - 10.1093/mnras/sty3497

M3 - Article

AN - SCOPUS:85062264285

SN - 0035-8711

VL - 484

SP - 93

EP - 100

JO - Monthly Notices of the Royal Astronomical Society

JF - Monthly Notices of the Royal Astronomical Society

IS - 1

ER -

Transfer learning for galaxy morphology from one survey to another

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this