Simulation-Based Inference of Strong Gravitational Lensing Parameters

Abstract

In the coming years, a new generation of sky surveys, in particular, Euclid Space Telescope (2022), and the Rubin Observatory's Legacy Survey of Space and Time (LSST, 2023) will discover more than 200,000 new strong gravitational lenses, which represents an increase of more than two orders of magnitude compared to currently known sample sizes [1]. Accurate and fast analysis of such large volumes of data under a statistical framework is therefore crucial for all sciences enabled by strong lensing. Here, we report on the application of simulation-based inference methods, in particular, density estimation techniques, to the predictions of the set of parameters of strong lensing systems from neural networks. This allows us to explicitly impose desired priors on lensing parameters, while guaranteeing convergence to the optimal posterior in the limit of perfect performance.

A Framework for Obtaining Accurate Posteriors of Strong Gravitational Lensing Parameters with Flexible Priors and Implicit Likelihoods Using Density Estimation

“…Our simulations consist of an SIE foreground galaxy with added external shear and an elliptical Sérsic background source. As in previous works (e.g., Hezaveh et al 2017;Perreault Levasseur et al 2017;Legin et al 2021Legin et al , 2022Schuldt et al 2022aSchuldt et al , 2022b, we use Cartesian coordinates (ò x , ò y ) for the ellipticity of both foreground galaxy and background source. This is a more suitable choice of parameters compared to ellipticity and orientation angle, since for more circular morphologies, the orientation angle becomes effectively unconstrained.…”

Machine Learning and Cosmology

“…Further work [10][11][12] proposed applying parameter inference and uncertainty quantification methods in order to characterize the properties of lensed sources and lensing galaxies. More recently, with an eye towards the large sample of gravitational lenses that will be imaged by forthcoming cosmological surveys like Euclid and LSST, there has been significant effort towards understanding how to utilize machine learning to optimally exploit this data towards source/lens characterization [13][14][15][16], Hubble constant inference [17], and characterization of dark matter substructure within the lensing galaxies [18][19][20][21][22][23][24][25][26][27][28][29] in a scalable manner.…”

Calibrating Cosmological Simulations with Implicit Likelihood Inference Using Galaxy Growth Observables

“…In conventional inference, although it generally depends on the complexity of the problem, the convergent MCMC typically requires at least 10 5 samples in cosmological applications (Feroz & Hobson 2008;Trotta et al 2011), which entails significantly more simulations than the NDE requires. ILI has already been vigorously exploited for inference and estimation of physical quantities in astrophysics, for example for inference of the Hubble constant from binary neutron star mergers (Gerardi et al 2021), constraints on the cosmological parameters from weak lensing (Tam et al 2022), mass estimations of the Milky Way and M31 (Lemos et al 2021;Villanueva-Domingo et al 2021), inference of strong gravitational lensing parameters (Legin et al 2021), dynamical mass estimation of galaxy clusters (Kodi Ramanah et al 2021), and inference of reionization parameters from the 21 cm power spectrum and light cones (Zhao et al 2022a(Zhao et al , 2022b.…”