Strong Lensing Source Reconstruction Using Continuous Neural Fields

Modeling galaxy-galaxy strong lensing observations A particular challenge with fully exploiting observations presents a number of challenges as the exact configuration of galaxy-galaxy strong gravitational lenses -- where an of both the background source and foreground extended background source is lensed by a foreground lens galaxy is unknown. A timely call, galaxy -- is that of accounting for the complex morphologies prompted by a number of upcoming surveys anticipating of lensed galaxies. Although sources in low-resolution high-resolution lensing images, demands images can be adequately modeled using phenomenological methods that can efficiently model lenses at their parameterizations such as one or several Sérsic profiles full complexity. In this work, we introduce a (Sérsic, 1963), this approach is inadequate for modeling method that uses continuous neural fields to nonparametrically higher-fidelity lensing observations such as those from ongoing, reconstruct the complex morphology upcoming, and proposed telescopes like the Hubble of a source galaxy while simultaneously inferring Space Telescope (HST), JWST, Euclid, and the Extremely a distribution over foreground lens galaxy Large Telescope (ELT). The development of new methods is configurations. We demonstrate the efficacy of especially timely, given the large number of high-resolution our method through experiments on simulated lenses that are expected to be imaged by next-generation data targeting high-resolution lensing images similar cosmological surveys (Collett, 2015) and their potential to to those anticipated in near-future astrophysical weigh in on the nature of dark matter (Simon et al., 2019).