CosmicRIM : Reconstructing Early Universe by Combining Differentiable Simulations with Recurrent Inference Machines

Abstract

Reconstructing the Gaussian initial conditions at the beginning of the Universe from the survey data in a forward modeling framework is a major challenge in cosmology. This requires solving a high dimensional inverse problem with an expensive, non-linear forward model: a cosmological N-body simulation. While intractable until recently, we propose to solve this inference problem using an automatically differentiable N-body solver, combined with a recurrent networks to learn the inference scheme and obtain the maximum-a-posteriori (MAP) estimate of the initial conditions of the Universe. We demonstrate using realistic cosmological observables that learnt inference is 40 times faster than traditional algorithms such as ADAM and LBFGS, which require specialized annealing schemes, and obtains solution of higher quality.

arXiv commentPublished as a workshop paper at ICLR 2021 SimDL Workshop

References in corpus (8)

• Adam: A Method for Stochastic Optimization
• Learning to learn by gradient descent by gradient descent
• Bayesian physical reconstruction of initial conditions from large scale structure surveys
• The clustering of galaxies in the completed SDSS-III Baryon Oscillation Spectroscopic Survey: Baryon Acoustic Oscillations in Fourier-space
• FastPM: a new scheme for fast simulations of dark matter and halos
• Modeling Biased Tracers at the Field Level
• Recurrent Inference Machines for Solving Inverse Problems
• Data-Driven Reconstruction of Gravitationally Lensed Galaxies using Recurrent Inference Machines

Cited by in corpus (4)

• Reconstructing the Universe with Variational self-Boosted Sampling
• Modeling iterative reconstruction and displacement field in the large scale structure
• Analytic auto-differentiable $\Lambda$CDM cosmography
• Fast Sampling of Cosmological Initial Conditions with Gaussian Neural Posterior Estimation