BibTeX

@article{2511.04681v1,
Author        = {A. Thomsen and J. Bucko and T. Kacprzak and V. Ajani and J. Fluri and A. Refregier and D. Anbajagane and F. J. Castander and A. Ferté and M. Gatti and N. Jeffrey and A. Alarcon and A. Amon and K. Bechtol and M. R. Becker and G. M. Bernstein and A. Campos and A. Carnero Rosell and C. Chang and R. Chen and A. Choi and M. Crocce and C. Davis and J. DeRose and S. Dodelson and C. Doux and K. Eckert and J. Elvin-Poole and S. Everett and P. Fosalba and D. Gruen and I. Harrison and K. Herner and E. M. Huff and M. Jarvis and N. Kuropatkin and P. -F. Leget and N. MacCrann and J. McCullough and J. Myles and A. Navarro-Alsina and S. Pandey and A. Porredon and J. Prat and M. Raveri and M. Rodriguez-Monroy and R. P. Rollins and A. Roodman and E. S. Rykoff and C. Sánchez and L. F. Secco and E. Sheldon and T. Shin and M. A. Troxel and I. Tutusaus and T. N. Varga and N. Weaverdyck and R. H. Wechsler and B. Yanny and B. Yin and Y. Zhang and J. Zuntz and S. Allam and F. Andrade-Oliveira and D. Bacon and J. Blazek and D. Brooks and R. Camilleri and J. Carretero and R. Cawthon and L. N. da Costa and M. E. da Silva Pereira and T. M. Davis and J. De Vicente and S. Desai and P. Doel and J. García-Bellido and G. Gutierrez and S. R. Hinton and D. L. Hollowood and K. Honscheid and D. J. James and K. Kuehn and O. Lahav and S. Lee and J. L. Marshall and J. Mena-Fernández and F. Menanteau and R. Miquel and J. Muir and R. L. C. Ogando and A. A. Plazas Malagón and E. Sanchez and D. Sanchez Cid and I. Sevilla-Noarbe and M. Smith and E. Suchyta and M. E. C. Swanson and D. Thomas and C. To and D. L. Tucker},
Title         = {Dark Energy Survey Year 3 results: Simulation-based $w$CDM inference from weak lensing and galaxy clustering maps with deep learning. I. Analysis design},
Eprint        = {2511.04681v1},
ArchivePrefix = {arXiv},
PrimaryClass  = {astro-ph.CO},
Abstract      = {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 CosmoGridV1 suite of N-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 $w$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 Buzzard galaxy catalogs. Our forecasts yield significant improvements in cosmological parameter constraints, achieving $2-3\times$ higher figures of merit in the $Ω_m - S_8$ 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.},
Year          = {2025},
Month         = {Nov},
Url           = {http://arxiv.org/abs/2511.04681v1},
File          = {2511.04681v1.pdf}
}