Abstract
Weak gravitational lensing is the slight distortion of galaxy shapes causedprimarily by the gravitational effects of dark matter in the universe. In ourwork, we seek to invert the weak lensing signal from 2D telescope images toreconstruct a 3D map of the universe's dark matter field. While inversiontypically yields a 2D projection of the dark matter field, accurate 3D maps ofthe dark matter distribution are essential for localizing structures ofinterest and testing theories of our universe. However, 3D inversion posessignificant challenges. First, unlike standard 3D reconstruction that relies onmultiple viewpoints, in this case, images are only observed from a singleviewpoint. This challenge can be partially addressed by observing how galaxyemitters throughout the volume are lensed. However, this leads to the secondchallenge: the shapes and exact locations of unlensed galaxies are unknown, andcan only be estimated with a very large degree of uncertainty. This introducesan overwhelming amount of noise which nearly drowns out the lensing signalcompletely. Previous approaches tackle this by imposing strong assumptionsabout the structures in the volume. We instead propose a methodology using agravitationally-constrained neural field to flexibly model the continuousmatter distribution. We take an analysis-by-synthesis approach, optimizing theweights of the neural network through a fully differentiable physical forwardmodel to reproduce the lensing signal present in image measurements. Weshowcase our method on simulations, including realistic simulated measurementsof dark matter distributions that mimic data from upcoming telescope surveys.Our results show that our method can not only outperform previous methods, butimportantly is also able to recover potentially surprising dark matterstructures.