LensFlow: A Convolutional Neural Network in Search of Strong Gravitational Lenses

Pourrahmani, Milad; Nayyeri, Hooshang; Cooray, Asantha

doi:10.3847/1538-4357/aaae6a

Cited by 70 publications

(59 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To produce the data used to train and validate the ConvNets, we adopt a hybrid approach similarly as done in Petrillo et al (2017); Jacobs et al (2017); Pourrahmani et al (2018), creating mock images of strong gravitational 4 https://github.com/drphilmarshall/HumVI lenses using images of real galaxies from KiDS and superimposing simulated lensed images. We adopt this approach because we do not have a sample of genuine KiDS lenses large enough to train a ConvNet (usually of the order of 10 6 ).…”

Section: Training the Convnets To Find Strong Lensesmentioning

confidence: 99%

Testing Convolutional Neural Networks for finding strong gravitational lenses in KiDS

Petrillo

Tortora

Chatterjee

et al. 2018

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

Convolutional Neural Networks (ConvNets) are one of the most promising methods for identifying strong gravitational lens candidates in survey data. We present two ConvNet lens-finders which we have trained with a dataset composed of real galaxies from the Kilo Degree Survey (KiDS) and simulated lensed sources. One ConvNet is trained with single r -band galaxy images, hence basing the classification mostly on the morphology. While the other ConvNet is trained on g-r-i composite images, relying mostly on colours and morphology. We have tested the ConvNet lens-finders on a sample of 21789 Luminous Red Galaxies (LRGs) selected from KiDS and we have analyzed and compared the results with our previous ConvNet lens-finder on the same sample. The new lens-finders achieve a higher accuracy and completeness in identifying gravitational lens candidates, especially the single-band ConvNet. Our analysis indicates that this is mainly due to improved simulations of the lensed sources. In particular, the single-band ConvNet can select a sample of lens candidates with ∼ 40% purity, retrieving 3 out of 4 of the confirmed gravitational lenses in the LRG sample. With this particular setup and limited human intervention, it will be possible to retrieve, in future surveys such as Euclid, a sample of lenses exceeding in size the total number of currently known gravitational lenses.

show abstract

Section: Training the Convnets To Find Strong Lensesmentioning

confidence: 99%

Testing Convolutional Neural Networks for finding strong gravitational lenses in KiDS

Petrillo

Tortora

Chatterjee

et al. 2018

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

show abstract

“…However, since there are only a few hundred unique confirmed lenses known to science, limiting training examples to real lens images is not the most effective approach to training a convolutional neural network. A common method of solving the problem of limited real training data is to simulate gravitational lenses using ray-tracing software (Jacobs et al 2017;Petrillo et al 2017;Pourrahmani et al 2018). We adopt a similar approach to that used by Pourrahmani et al (2018).…”

Section: Creating Simulated Images Of Gravitational Lensesmentioning

confidence: 99%

Comparison of Multi-class and Binary Classification Machine Learning Models in Identifying Strong Gravitational Lenses

Teimoorinia

Toyonaga

Fabbro

et al. 2020

PASP

View full text Add to dashboard Cite

Typically, binary classification lens-finding schemes are used to discriminate between lens candidates and non-lenses. However, these models often suffer from substantial false-positive classifications. Such false positives frequently occur due to images containing objects such as crowded sources, galaxies with arms, and also images with a central source and smaller surrounding sources. Therefore, a model might confuse the stated circumstances with an Einstein ring. It has been proposed that by allowing such commonly misclassified image types to constitute their own classes, machine learning models will more easily be able to learn the difference between images that contain real lenses, and images that contain lens imposters. Using Hubble Space Telescope (HST) images, in the F814W filter, we compare the usage of binary and multi-class classification models applied to the lens finding task. From our findings, we conclude there is not a significant benefit to using the multi-class model over a binary model. We will also present the results of a simple lens search using a multi-class machine learning model, and potential new lens candidates.

show abstract

“…Conscious of the unique opportunity brought by these modern large sky surveys, numerous methods were recently developed to systematically search for GLs (Bolton et al 2008;More et al 2016;Paraficz et al 2016;Agnello et al 2018a,b;Jacobs et al 2017;Lee 2017;Pourrahmani et al 2018;Lemon et al 2019). At the state of the art of these identification techniques, the Strong Gravitational Lens Finding Challenge (Metcalf et al 2018) is a recent effort to identify GLs in large scale imaging surveys such as the upcoming Square Kilometer Array (SKA) 1 , the Large Synoptic Survey Telescope (LSST) 2 , and the Euclid space telescope 3 .…”

Section: Introductionmentioning

confidence: 99%

Gaia GraL: Gaia DR2 Gravitational Lens Systems

Delchambre

Krone-Martins

Wertz

et al. 2019

A&A

View full text Add to dashboard Cite

Aims. In this work, we aim to provide a reliable list of gravitational lens candidates based on a search performed over the entire Gaia Data Release 2 (Gaia DR2). We also aim to show that the astrometric and photometric information coming from the Gaia satellite yield sufficient insights for supervised learning methods to automatically identify strong gravitational lens candidates with an efficiency that is comparable to methods based on image processing. Methods. We simulated 106 623 188 lens systems composed of more than two images, based on a regular grid of parameters characterizing a non-singular isothermal ellipsoid lens model in the presence of an external shear. These simulations are used as an input for training and testing our supervised learning models consisting of extremely randomized trees (ERTs). These trees are finally used to assign to each of the 2 129 659 clusters of celestial objects extracted from the Gaia DR2 a discriminant value that reflects the ability of our simulations to match the observed relative positions and fluxes from each cluster. Once complemented with additional constraints, these discriminant values allow us to identify strong gravitational lens candidates out of the list of clusters. Results. We report the discovery of 15 new quadruply-imaged lens candidates with angular separations of less than 6 and assess the performance of our approach by recovering 12 of the 13 known quadruply-imaged systems with all their components detected in Gaia DR2 with a misclassification rate of fortuitous clusters of stars as lens systems that is below 1%. Similarly, the identification capability of our method regarding quadruply-imaged systems where three images are detected in Gaia DR2 is assessed by recovering 10 of the 13 known quadruply-imaged systems having one of their constituting images discarded. The associated misclassification rate varies between 5.83% and 20%, depending on the image we decided to remove.

show abstract

LensFlow: A Convolutional Neural Network in Search of Strong Gravitational Lenses

Cited by 70 publications

References 61 publications

Testing Convolutional Neural Networks for finding strong gravitational lenses in KiDS

Testing Convolutional Neural Networks for finding strong gravitational lenses in KiDS

Comparison of Multi-class and Binary Classification Machine Learning Models in Identifying Strong Gravitational Lenses

Gaia GraL: Gaia DR2 Gravitational Lens Systems

Contact Info

Product

Resources

About