Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI
arXiv:2507.00514
Abstract
The simulation cost for cosmological simulation-based inference can be decreased by combining simulation sets of varying fidelity. We propose an approach to such multi-fidelity inference based on feature matching and knowledge distillation. Our method results in improved posterior quality, particularly for small simulation budgets and difficult inference problems.
5 pages, 4 figures; accepted at ICML-colocated ML4Astro 2025 workshop