Interpreting Cosmological Information from Neural Networks in the Hydrodynamic Universe

Abstract

What happens when a black box (neural network) meets a black box (simulation of the Universe)? Recent work has shown that convolutional neural networks (CNNs) can infer cosmological parameters from the matter density field in the presence of complex baryonic processes. A key question that arises is, which parts of the cosmic web is the neural network obtaining information from? We shed light on the matter by identifying the Fourier scales, density scales, and morphological features of the cosmic web that CNNs pay most attention to. We find that CNNs extract cosmological information from both high and low density regions: overdense regions provide the most information per pixel, while underdense regions -- particularly deep voids and their surroundings -- contribute significantly due to their large spatial extent and coherent spatial features. Remarkably, we demonstrate that there is negligible degradation in cosmological constraining power after aggressive cutting in both maximum Fourier scale and density. Furthermore, we find similar results when considering both hydrodynamic and gravity-only simulations, implying that neural networks can marginalize over baryonic effects with minimal loss in cosmological constraining power. Our findings point to practical strategies for optimal and robust field-level cosmological inference in the presence of uncertainly modeled astrophysics.

14 pages, 11 figures, 1 table

References in corpus (33)

• First results from the IllustrisTNG simulations: the stellar mass content of groups and clusters of galaxies
• The Arepo public code release
• The Neutrino Puzzle: Anomalies, Interactions, and Cosmological Tensions
• Bayesian physical reconstruction of initial conditions from large scale structure surveys
• The CAMELS project: Cosmology and Astrophysics with MachinE Learning Simulations
• A non-linear solution to the $S_8$ tension?
• Rejuvenating the matter power spectrum: restoring information with a logarithmic density mapping
• Physical Bayesian modelling of the non-linear matter distribution: new insights into the Nearby Universe
• Towards optimal extraction of cosmological information from nonlinear data
• Detecting Neutrino Mass by Combining Matter Clustering, Halos, and Voids
• A rigorous EFT-based forward model for large-scale structure
• The GIGANTES dataset: precision cosmology from voids in the machine learning era
• The CAMELS Multifield Dataset: Learning the Universe's Fundamental Parameters with Artificial Intelligence
• Machine learning cosmological structure formation
• How much information can be extracted from galaxy clustering at the field level?
• Translation and Rotation Equivariant Normalizing Flow (TRENF) for Optimal Cosmological Analysis
• Clipping the Cosmos: The Bias and Bispectrum of Large Scale Structure
• Interpreting deep learning models for weak lensing
• Unbiased Cosmology Inference from Biased Tracers using the EFT Likelihood
• An interpretable machine learning framework for dark matter halo formation
• Consistency tests of field level inference with the EFT likelihood
• Beware of Fake $\nu$s: The Effect of Massive Neutrinos on the Nonlinear Evolution of Cosmic Structure
• Impacts of the physical data model on the forward inference of initial conditions from biased tracers
• Baryon acoustic oscillations reconstruction using convolutional neural networks
• Explaining deep learning of galaxy morphology with saliency mapping
• Joint velocity and density reconstruction of the Universe with nonlinear differentiable forward modeling
• A comparative study of cosmological constraints from weak lensing using Convolutional Neural Networks
• Explaining dark matter halo density profiles with neural networks
• Deep learning insights into cosmological structure formation
• Significance of void shape: Neutrino mass from Voronoi void halos?
• Massive $\nu$s through the CNN lens: interpreting the field-level neutrino mass information in weak lensing
• Differentiable Cosmological Hydrodynamics for Field-Level Inference and High Dimensional Parameter Constraints
• The Power of the Cosmic Web

Cited by in corpus (1)

• Interpretability of deep-learning methods applied to large-scale structure surveys