Gradients of the Big Bang with Bolt.jl

Zack Li (Canadian Institute for Theoretical Astrophysics / University of Toronto)

Bolt.jl performs differentiable simulations of the early Universe using linear perturbation theory. The results of these simulations can be compared with astrophysical observations to extract cutting-edge measurements of the age, curvature, and contents of the Universe. In this talk, we highlight how Julia uniquely enables the differentiability of these complex solvers, and discuss how gradients and GPU support can accelerate cosmological parameter estimation and model optimization.