Benchmarking Exact, GP-Emulated, and Simulation-Based Inference for Late-Time Cosmology
Abstract
Forthcoming cosmological surveys require inference pipelines that are both statistically reliable and computationally scalable. In this work, we perform a systematic comparison of three complementary inference strategies for late-time ΛCDM cosmology: exact Markov Chain Monte Carlo (MCMC), Gaussian Process (GP)-assisted MCMC, and neural Simulation-Based Inference (SBI). Using a common analysis framework based on Cosmic Chronometers, DESI DR2 baryon acoustic oscillation measurements, and the Pantheon+ Type Ia supernova compilation, we consider two dataset combinations of increasing complexity, namely CC+DESI and CC+DESI+PP, under identical cosmological assumptions and priors. For CC+DESI, both GP emulation and SBI reproduce the exact posterior constraints on (H0,Ωm,0) to better than 0.3σ. For the more constraining CC+DESI+PP combination, modest method-dependent shifts emerge, reaching at most ∼1.5σ in a single parameter. Despite these differences, all methods recover a nearly identical expansion history, with percent-level agreement across the full redshift range. From a computational perspective, GP emulation accelerates model evaluations but remains limited by MCMC sampling, whereas SBI achieves order-of-magnitude reductions in total runtime through amortized posterior learning. We further investigate the convergence of SBI as a function of simulation budget and identify the number of simulations required to obtain stable posterior constraints. Overall, our results demonstrate that accelerated inference techniques can deliver reliable cosmological constraints for realistic late-time datasets considered here at a fraction of the computational cost of conventional likelihood-based analyses.
Versions
Related
- Benchmarking Exact, Gaussian Process-emulated, and SBI for Late-time Cosmology
- Merlin: Fast and flexible 3x2pt cosmology with simulation-based inference
- Field-level weak lensing cosmology with $60$ simulations using multifidelity simulation-based inference
- FPCA-Enhanced Simulation-Based Inference for Robust Type Ia Supernova Cosmology
- 2Flow2SN: Dusty Supernova Cosmology with Photometric Redshifts and Contamination using Neural Simulation-Based Inference
BibTeX
@misc{mishra2026benchmarking,
title = {{Benchmarking Exact, GP-Emulated, and Simulation-Based Inference for Late-Time Cosmology}},
author = {Sai Swagat Mishra},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2606.16248},
eprint = {2606.16248},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.16248},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Benchmarking Exact, GP-Emulated, and Simulation-Based Inference for Late-Time Cosmology
AU - Sai Swagat Mishra
PY - 2026
DA - 2026/09/17
PB - arXiv
DO - 10.48550/arxiv.2606.16248
UR - https://arxiv.org/abs/2606.16248
AB - Forthcoming cosmological surveys require inference pipelines that are both statistically reliable and computationally scalable. In this work, we perform a systematic comparison of three complementary inference strategies for late-time ΛCDM cosmology: exact Markov Chain Monte Carlo (MCMC), Gaussian Process (GP)-assisted MCMC, and neural Simulation-Based Inference (SBI). Using a common analysis framework based on Cosmic Chronometers, DESI DR2 baryon acoustic oscillation measurements, and the Pantheon+ Type Ia supernova compilation, we consider two dataset combinations of increasing complexity, namely CC+DESI and CC+DESI+PP, under identical cosmological assumptions and priors. For CC+DESI, both GP emulation and SBI reproduce the exact posterior constraints on (H_0,Ω_m,0) to better than 0.3σ. For the more constraining CC+DESI+PP combination, modest method-dependent shifts emerge, reaching at most ∼1.5σ in a single parameter. Despite these differences, all methods recover a nearly identical expansion history, with percent-level agreement across the full redshift range. From a computational perspective, GP emulation accelerates model evaluations but remains limited by MCMC sampling, whereas SBI achieves order-of-magnitude reductions in total runtime through amortized posterior learning. We further investigate the convergence of SBI as a function of simulation budget and identify the number of simulations required to obtain stable posterior constraints. Overall, our results demonstrate that accelerated inference techniques can deliver reliable cosmological constraints for realistic late-time datasets considered here at a fraction of the computational cost of conventional likelihood-based analyses.
ER -