<i>Euclid</i>: Searches for strong gravitational lenses using convolutional neural nets in Early Release Observations of the Perseus field

The Euclid Wide Survey (EWS) is predicted to find approximately 170 000 galaxy-galaxy strong lenses from its lifetime observation of 14 000 deg 2 of the sky. Detecting this many lenses by visual inspection with professional astronomers and citizen scientists alone is infeasible. As a result, machine learning algorithms, particularly convolutional neural networks (CNNs), have been used as an automated method of detecting strong lenses, and have proven fruitful in finding galaxy-galaxy strong lens candidates, such that the usage of CNNs in lens identification has increased. We identify the major challenge to be the automatic detection of galaxy-galaxy strong lenses while simultaneously maintaining a low false positive rate, thus producing a pure and complete sample of strong lens candidates from Euclid with a limited need for visual inspection. One aim of this research is to have a quantified starting point on the achieved purity and completeness with our current version of CNN-based detection pipelines for the VIS images of EWS. This work is vital in preparing our CNN-based detection pipelines to be able to produce a pure sample of the &gt;100 000 strong gravitational lensing systems widely predicted for Euclid . We select all sources with VIS I E &lt; 23 mag from the Euclid Early Release Observation imaging of the Perseus field. We apply a range of CNN architectures to detect strong lenses in these cutouts. All our networks perform extremely well on simulated data sets and their respective validation sets. However, when applied to real Euclid imaging, the highest lens purity is just ∼11%. Among all our networks, the false positives are typically identifiable by human volunteers as, for example, spiral galaxies, multiple sources, and artifacts, implying that improvements are still possible, perhaps via a second, more interpretable lens selection filtering stage. There is currently no alternative to human classification of CNN-selected lens candidates. Given the expected ∼10 5 lensing systems in Euclid , this implies 10 6 objects for human classification, which while very large is not in principle intractable and not without precedent.

Authors

Institutions

Publication Details

Journal
Astronomy and Astrophysics
Published
2025-03-27
DOI
https://doi.org/10.1051/0004-6361/202453152
Citations
10
Primary Topic
Advanced Measurement and Metrology Techniques
Type
article
Field-Weighted Citation Impact
7.13

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Euclid: Searches for strong gravitational lenses using convolutional neural nets in Early Release Observations of the Perseus field

Predrag Matavulj, A. Hornstrup, S. Serjeant, I.T Andika et al.
10 citations
Astronomy and Astrophysics
Advanced Measurement and Metrology Techniques
7.13
article

Euclid: Searches for strong gravitational lenses using convolutional neural nets in Early Release Observations of the Perseus field

Predrag Matavulj, A. Hornstrup, S. Serjeant, I.T Andika, Anna G. Manjón, B.C Nagam, Valerio Busillo, R Pearce-Casey, L. Ulivi, M. Walmsley, L. Leuzzi, de Wilde Jaap
article en
10 citations

Abstract

The Euclid Wide Survey (EWS) is predicted to find approximately 170 000 galaxy-galaxy strong lenses from its lifetime observation of 14 000 deg 2 of the sky. Detecting this many lenses by visual inspection with professional astronomers and citizen scientists alone is infeasible. As a result, machine learning algorithms, particularly convolutional neural networks (CNNs), have been used as an automated method of detecting strong lenses, and have proven fruitful in finding galaxy-galaxy strong lens candidates, such that the usage of CNNs in lens identification has increased. We identify the major challenge to be the automatic detection of galaxy-galaxy strong lenses while simultaneously maintaining a low false positive rate, thus producing a pure and complete sample of strong lens candidates from Euclid with a limited need for visual inspection. One aim of this research is to have a quantified starting point on the achieved purity and completeness with our current version of CNN-based detection pipelines for the VIS images of EWS. This work is vital in preparing our CNN-based detection pipelines to be able to produce a pure sample of the &gt;100 000 strong gravitational lensing systems widely predicted for Euclid . We select all sources with VIS I E &lt; 23 mag from the Euclid Early Release Observation imaging of the Perseus field. We apply a range of CNN architectures to detect strong lenses in these cutouts. All our networks perform extremely well on simulated data sets and their respective validation sets. However, when applied to real Euclid imaging, the highest lens purity is just ∼11%. Among all our networks, the false positives are typically identifiable by human volunteers as, for example, spiral galaxies, multiple sources, and artifacts, implying that improvements are still possible, perhaps via a second, more interpretable lens selection filtering stage. There is currently no alternative to human classification of CNN-selected lens candidates. Given the expected ∼10 5 lensing systems in Euclid , this implies 10 6 objects for human classification, which while very large is not in principle intractable and not without precedent.

Astronomy and AstrophysicsVol. 696
Université Claude Bernard Lyon 1 (FR), University of Geneva (CH), Institució Catalana de Recerca i Estudis Avançats (ES), California Institute of Technology (US), University of Copenhagen (DK), Centre National de la Recherche Scientifique (FR), University of Minnesota (US), Goddard Space Flight Center (US), Institute of Space Sciences (ES), Jet Propulsion Laboratory (US), University of Helsinki (FI), American Museum of Natural History (US), Université Toulouse III - Paul Sabatier (FR), University of Bonn (DE), Scuola Internazionale Superiore di Studi Avanzati (IT), University of Padua (IT), University of Lisbon (PT), University of Trieste (IT), University of Salento (IT), Université de Bordeaux (FR), University of Waterloo (CA), University of Cape Town (ZA), University of Liège (BE), Lehman College (US), University of Groningen (NL), Université Fédérale de Toulouse Midi-Pyrénées (FR), Université Catholique de Lille (FR), Universidade do Porto (PT), Perimeter Institute (CA), Saint Joseph University (LB), University of Oslo (NO), University of Toronto (CA), Institute for High Energy Physics (ES), University of Milan (IT), Durham University (GB), Universität Innsbruck (AT), University of Trento (IT), University of Ferrara (IT), The Open University (GB), Université Paris Cité (FR), Aix-Marseille Université (FR), Heidelberg University (DE), Commissariat à l'Énergie Atomique et aux Énergies Alternatives (FR), Université Paris-Saclay (FR), Centre National d'Études Spatiales (FR), Arcetri Astrophysical Observatory (IT), Helsinki Institute of Physics (FI), Astronomical Observatory of Rome (IT), Agenzia Spaziale Italiana (IT), Trieste Astronomical Observatory (IT), Institut d'Astrophysique de Paris (FR), Astronomical Observatory of Capodimonte (IT), Laboratoire AstroParticule et Cosmologie (FR), Observatoire astronomique de Strasbourg (FR), Brera Astronomical Observatory (IT), University of Surrey (GB), University of Manchester (GB), FHNW University of Applied Sciences and Arts Northwestern Switzerland (CH), Institute for High Performance Computing and Networking (IT), Universidad Politécnica de Cartagena (ES), Sorbonne Université (FR), University of Oxford (GB), Técnicas Reunidas (Spain) (ES), Astrophysique, Instrumentation et Modélisation (FR), Institute of Astrophysics and Space Sciences (PT), Max Planck Institute for Extraterrestrial Physics (DE), Institut d'Estudis Espacials de Catalunya (ES), European Space Astronomy Centre (ES), Istituto Nazionale di Fisica Nucleare, Sezione di Lecce (IT), Laboratoire Univers et Particules de Montpellier (FR), Laboratoire d'Astrophysique de Bordeaux (FR), Istituto Nazionale di Fisica Nucleare, Sezione di Torino (IT), Institute for Space Astrophysics and Planetology (IT), Max Planck Institute for Astrophysics (DE), Istituto Nazionale di Fisica Nucleare, Sezione di Padova (IT), Osservatorio Astrofisico di Torino (IT), Royal Observatory (GB), Istituto Nazionale di Fisica Nucleare, Sezione di Napoli (IT), Centro de Astrofísica da Universidade do Porto (PT), Max Planck Institute for Astronomy (DE), Istituto di Radioastronomia di Bologna (IT), Cegep Edouard Montpetit (CA), Istituto Nazionale di Fisica Nucleare, Sezione di Genova (IT), CEA Paris-Saclay (FR), Instituto de Astrofísica de Canarias (ES), Institut National de Physique Nucléaire et de Physique des Particules (FR), Institute of Space Science - INFLPR Subsidiary (RO), Osservatorio Astronomico della Regione Autonoma Valle d'Aosta (IT), Istituto Nazionale di Fisica Nucleare, Sezione di Bologna (IT), Port d'Informació Científica (ES), Istituto Nazionale di Fisica Nucleare, Sezione di Trieste (IT), Institute for Fundamental Physics of the Universe (IT), Damien Foundation (BD), Istituto Nazionale di Fisica Nucleare, Sezione di Roma I (IT), Osservatorio Astronomico di Padova (IT), European Space Research Institute (IT), Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas (ES), Institut de Recherche en Astrophysique et Planétologie (FR), Infrared Processing and Analysis Center (US), Osservatorio di Astrofisica e Scienza dello Spazio, Fondazione ICSC Centro Nazionale di Ricerca in High Performance Computing, Big Data e Quantum Computing, Inter-university Institute for Data Intensive Astronomy (ZA), European Space Research and Technology Centre (NL), University of Florence (IT), University College London (GB), Space Research Organisation Netherlands (NL), École Polytechnique Fédérale de Lausanne (CH), University of Turin (IT), Technical University of Munich (DE), University of Portsmouth (GB), Lancaster University (GB), Université de Strasbourg (FR), University of Chile (CL), Heidelberg Institute for Theoretical Studies (DE), Universitat de Barcelona (ES), Ludwig-Maximilians-Universität München (DE), University of Genoa (IT), Newcastle University (GB), University of the Western Cape (ZA), National Institute for Astrophysics (IT), RWTH Aachen University (DE), Netherlands Institute for Radio Astronomy (NL), University of Bologna (IT), Universitat Politècnica de Catalunya (ES), Technical University of Denmark (DK), University of Edinburgh (GB)
National Aeronautics and Space Administration, Ministerio de Ciencia, Innovación y Universidades, Department of Science and Innovation, South Africa, European Space Agency, European Commission, National Research Foundation, Deutsche Forschungsgemeinschaft, University of Toronto, Magyar Tudományos Akadémia, Agenzia Spaziale Italiana, Max-Planck-Gesellschaft, National Astronomical Observatory of Japan, Agenția Spațială Română, Norsk Romsenter, Science and Technology Facilities Council, Fundação para a Ciência e a Tecnologia, Agencia Estatal de Investigación
Partnerships for the goals
Openalex Percentile: Top 3%
Advanced Measurement and Metrology Techniques
7.13
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.