Euclid preparation. LIII. LensMC, weak lensing cosmic shear measurement with forward modelling and Markov Chain Monte Carlo sampling
Authors:
Euclid Collaboration,
G. Congedo,
L. Miller,
A. N. Taylor,
N. Cross,
C. A. J. Duncan,
T. Kitching,
N. Martinet,
S. Matthew,
T. Schrabback,
M. Tewes,
N. Welikala,
N. Aghanim,
A. Amara,
S. Andreon,
N. Auricchio,
M. Baldi,
S. Bardelli,
R. Bender,
C. Bodendorf,
D. Bonino,
E. Branchini,
M. Brescia,
J. Brinchmann,
S. Camera
, et al. (217 additional authors not shown)
Abstract:
LensMC is a weak lensing shear measurement method developed for Euclid and Stage-IV surveys. It is based on forward modelling in order to deal with convolution by a point spread function (PSF) with comparable size to many galaxies; sampling the posterior distribution of galaxy parameters via Markov Chain Monte Carlo; and marginalisation over nuisance parameters for each of the 1.5 billion galaxies…
▽ More
LensMC is a weak lensing shear measurement method developed for Euclid and Stage-IV surveys. It is based on forward modelling in order to deal with convolution by a point spread function (PSF) with comparable size to many galaxies; sampling the posterior distribution of galaxy parameters via Markov Chain Monte Carlo; and marginalisation over nuisance parameters for each of the 1.5 billion galaxies observed by Euclid. We quantified the scientific performance through high-fidelity images based on the Euclid Flagship simulations and emulation of the Euclid VIS images; realistic clustering with a mean surface number density of 250 arcmin$^{-2}$ ($I_{\rm E}<29.5$) for galaxies, and 6 arcmin$^{-2}$ ($I_{\rm E}<26$) for stars; and a diffraction-limited chromatic PSF with a full width at half maximum of $0.^{\!\prime\prime}2$ and spatial variation across the field of view. LensMC measured objects with a density of 90 arcmin$^{-2}$ ($I_{\rm E}<26.5$) in 4500 deg$^2$. The total shear bias was broken down into measurement (our main focus here) and selection effects (which will be addressed elsewhere). We found measurement multiplicative and additive biases of $m_1=(-3.6\pm0.2)\times10^{-3}$, $m_2=(-4.3\pm0.2)\times10^{-3}$, $c_1=(-1.78\pm0.03)\times10^{-4}$, $c_2=(0.09\pm0.03)\times10^{-4}$; a large detection bias with a multiplicative component of $1.2\times10^{-2}$ and an additive component of $-3\times10^{-4}$; and a measurement PSF leakage of $α_1=(-9\pm3)\times10^{-4}$ and $α_2=(2\pm3)\times10^{-4}$. When model bias is suppressed, the obtained measurement biases are close to Euclid requirement and largely dominated by undetected faint galaxies ($-5\times10^{-3}$). Although significant, model bias will be straightforward to calibrate given the weak sensitivity. LensMC is publicly available at https://gitlab.com/gcongedo/LensMC
△ Less
Submitted 2 December, 2024; v1 submitted 1 May, 2024;
originally announced May 2024.
Weak-lensing shear measurement with machine learning: teaching artificial neural networks about feature noise
Authors:
Malte Tewes,
Thibault Kuntzer,
Reiko Nakajima,
Frédéric Courbin,
Hendrik Hildebrandt,
Tim Schrabback
Abstract:
Cosmic shear is a primary cosmological probe for several present and upcoming surveys investigating dark matter and dark energy, such as Euclid or WFIRST. The probe requires an extremely accurate measurement of the shapes of millions of galaxies based on imaging data. Crucially, the shear measurement must address and compensate for a range of interwoven nuisance effects related to the instrument o…
▽ More
Cosmic shear is a primary cosmological probe for several present and upcoming surveys investigating dark matter and dark energy, such as Euclid or WFIRST. The probe requires an extremely accurate measurement of the shapes of millions of galaxies based on imaging data. Crucially, the shear measurement must address and compensate for a range of interwoven nuisance effects related to the instrument optics and detector, noise, unknown galaxy morphologies, colors, blending of sources, and selection effects. This paper explores the use of supervised machine learning (ML) as a tool to solve this inverse problem. We present a simple architecture that learns to regress shear point estimates and weights via shallow artificial neural networks. The networks are trained on simulations of the forward observing process, and take combinations of moments of the galaxy images as inputs. A challenging peculiarity of this ML application is the combination of the noisiness of the input features and the requirements on the accuracy of the inverse regression. To address this issue, the proposed training algorithm minimizes bias over multiple realizations of individual source galaxies, reducing the sensitivity to properties of the overall sample of source galaxies. Importantly, an observational selection function of these source galaxies can be straightforwardly taken into account via the weights. We first introduce key aspects of our approach using toy-model simulations, and then demonstrate its potential on images mimicking Euclid data. Finally, we analyze images from the GREAT3 challenge, obtaining competitively low shear biases despite the use of a simple training set. We conclude that the further development of ML approaches is of high interest to meet the stringent requirements on the shear measurement in current and future surveys. A demonstration implementation of our technique is publicly available.
△ Less
Submitted 30 January, 2019; v1 submitted 5 July, 2018;
originally announced July 2018.
