MADLens, a python package for fast and differentiable non-Gaussian lensing simulations

We present MADLens a python package for producing non-Gaussian lensing convergence maps at arbitrary source redshifts with unprecedented precision. is designed to achieve high accuracy while keeping computational costs as low possible. A simulation only 2563 particles produces whose power agrees theoretical spectra up L=10000 within the limits of HaloFit. This made possible by combination highly parallelizable particle-mesh algorithm, sub-evolution scheme in projection, and machine-learning inspired sharpening step. Further, fully differentiable respect initial conditions underlying simulations number cosmological parameters. These properties allow be used forward model Bayesian inference algorithms that require optimization or derivative-aided sampling. Another use case production large, resolution sets they are required training novel deep-learning-based analysis tools. make publicly available under Creative Commons License .