AI Maps the Universe's Hidden Matter to Unlock Cosmology's Biggest Mysteries
Physicists have deployed deep learning to analyze light-bending patterns across millions of galaxies, extracting far more information than traditional methods. The breakthrough could transform how scientists measure dark energy and the universe's composition—data worth billions to space agencies and fundamental to planning next-generation observatories.
Originaltitel: Dark Energy Survey Year 3 results: Simulation-based <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"> <mml:mrow> <mml:mi>w</mml:mi> <mml:mi>CDM</mml:mi> </mml:mrow> </mml:math> inference from weak lensing and galaxy clustering maps with deep learning: Analysis design
Data-driven approaches using deep learning are emerging as powerful techniques to extract non-Gaussian information from cosmological large-scale structure. This work presents the first simulation-based inference (SBI) pipeline that combines weak lensing and galaxy clustering maps in a realistic Dark Energy Survey Year 3 (DES Y3) configuration and serves as preparation for a forthcoming analysis of the survey data. We develop a scalable forward model based on the 1 suite of <a:math xmlns:a="http://www.w3.org/1998/Math/MathML" display="inline"> <a:mi>N</a:mi> </a:math> -body simulations to generate over one million self-consistent mock realizations of DES Y3 at the map level. Leveraging this large dataset, we train deep graph convolutional neural networks on the full survey footprint in spherical geometry to learn low-dimensional features that approximately maximize mutual information with target parameters. These learned compressions enable neural density estimation of the implicit likelihood via normalizing flows in a ten-dimensional parameter space spanning cosmological <c:math xmlns:c="http://www.w3.org/1998/Math/MathML" display="inline"> <c:mi>w</c:mi> </c:math> CDM, intrinsic alignment, and linear galaxy bias parameters, while marginalizing over baryonic, photometric redshift, and shear bias nuisances. To ensure robustness, we extensively validate our inference pipeline using synthetic observations derived from both systematic contaminations in our forward model and independent galaxy catalogs. Our forecasts yield significant improvements in cosmological parameter constraints, achieving <e:math xmlns:e="http://www.w3.org/1998/Math/MathML" display="inline"> <e:mn>2</e:mn> <e:mi>–</e:mi> <e:mn>3</e:mn> <e:mo>×</e:mo> </e:math> higher figures of merit in the <g:math xmlns:g="http://www.w3.org/1998/Math/MathML" display="inline"> <g:msub> <g:mi mathvariant="normal">Ω</g:mi> <g:mi>m</g:mi> </g:msub> <g:mtext>−</g:mtext> <g:msub> <g:mi>S</g:mi> <g:mn>8</g:mn> </g:msub> </g:math> plane relative to our implementation of baseline two-point statistics and effectively breaking parameter degeneracies through probe combination. These results demonstrate the potential of SBI analyses powered by deep learning for upcoming stage-IV wide-field imaging surveys.