Field-level weak lensing cosmology with $<100$ simulations using multifidelity simulation-based inference
2026-06-22 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors show a way to analyze cosmic weak lensing data using smart neural networks and simulations without needing a huge number of expensive, detailed simulations. They use a two-step process: first training on many quick, simplified simulations, then fine-tuning using fewer, more accurate ones. This approach captures more information from the data than traditional methods and cuts down the costly computations drastically. Their method works well with as few as 60 to 100 high-quality simulations to get trustworthy cosmological results.
weak lensingKiDS surveyN-body simulationsimulation-based inferenceneural compressioncosmological parameterspower spectrumfield-level inferencemultifidelity simulationslog-normal GLASS simulations
Authors
Alex A. Saoulis, Kiyam Lin, Niall Jeffrey, Maximilian von Wietersheim-Kramsta, Davide Piras, Alessio Spurio Mancini, Ana M. G. Ferreira, Benjamin Joachimi
Abstract
We perform a realistic KiDS-Legacy mock analysis with field-level neural compression and simulation-based inference using fewer than 100 $N$-body simulations. The weak lensing shear field encodes substantially more cosmological information than standard two-point summary statistics such as the power spectrum. Field-level inference can fully exploit this information, but physical realism at the field-level requires very high-fidelity simulations. This poses a major challenge for simulation-based inference (SBI): accurate empirical density modelling and deep-learning-based neural compression require many training simulations, but achieving physical realism at the field level makes each simulation extremely costly. We demonstrate that multifidelity SBI can alleviate this tension by substantially reducing the number of high-fidelity simulations needed for accurate cosmological inference. We pre-train neural inference models on realistic KiDS-Legacy-like shear mocks using fast log-normal GLASS simulations and fine-tune them on a small set of high-fidelity $N$-body simulations. We show that between $60$-$100$ high-fidelity simulations are sufficient to obtain informative and well-calibrated cosmological posteriors, enabling an order-of-magnitude reduction in simulation cost for accurate field-level inference in a realistic setting.