Field-level weak lensing cosmology with $<100$ simulations using multifidelity simulation-based inference
📄 arXiv:2606.23346 · 📥 PDF · 2026-06-22 · astro-ph.CO
Authors: Alex A. Saoulis [arXiv · scholar] , Kiyam Lin [arXiv · scholar] , Niall Jeffrey [arXiv · scholar] , Maximilian von Wietersheim-Kramsta [arXiv · scholar] , Davide Piras [arXiv · scholar] , Alessio Spurio Mancini [arXiv · scholar] , Ana M. G. Ferreira [arXiv · scholar] , Benjamin Joachimi [arXiv · scholar]
🕰 Orloj analysis
Tento článek představuje metodu pro kosmologickou inferenci slabého gravitačního čočkování na úrovni pole pomocí simulací a neuronové komprese. Využívá multifidelitní přístup k výraznému snížení počtu potřebných N-body simulací, čímž dosahuje úspory nákladů.
💡 Práce přináší významné praktické zlepšení v efektivitě kosmologické inference, umožňující řádové snížení výpočetních nákladů pro analýzy na úrovni pole.
Categories:
THE-1
COS-1
MET-2
MET-1
✓ falsifiable, limit_reductions, modest_claims
⚠ Reliance on mock data for validation, Potential fragility of neural compression with limited high-fidelity data
📄 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.