Reconstruction Strikes Back: Unveiling the Dynamics of the Universe in Next‐Generation Spectroscopic Surveys
IFPU
14-18 April 2025
Galaxies are discrete, biased tracers that provide a distorted snapshot of the underlying cosmological matter den‐ sity field. On large scales they outline an intricate network of structures known as the cosmic web. Understanding the evolution of these large‐scale structures helps us to study the composition of the Universe and to probe devi‐ ations from the standard cosmological model. To achieve this, we require precise reconstructions of the full, true, undistorted density and peculiar velocity fields both at the current time (Eulerian) and in the initial conditions (Lagrangian), constructed from the incomplete and biased information provided by galaxy redshift surveys. This Team Research week will bring together experts in Eulerian (PGV, AN) and Lagrangian (ES, FN, RKS) recon‐ struction methods to explore the strengths, limitations and synergies between them. The former exploits recent progress in Machine Learning approaches – Auto‐encoder Neural Networks and Generative Adversarial Networks – to recovering signals from noisy data. The latter exploits cutting edge advances in Optimal Transport, and more general variational methods. By the end of the week, we expect to have produced the first ever marriage of the Eulerian and Lagrangian approaches in a realistic cosmological simulation. The longer term goal is not just to re‐ construct, but to understand how the statistical uncertainties in each step of the reconstruction impact the final cosmological constraints. This effort is both crucial and timely, especially with the upcoming large surveys like EUCLID, DESI, LSST, SPHEREX and others on the horizon. It is also interdisciplinary: Both the machine learning and optimal transport methods have demonstrated potential for applications beyond cosmology.
Organisers:
- Punyakoti Ganeshaiah Veena (University of Genoa)
- Elena Sarpa (SISSA)
Participants:
- TBA