In plain words: Scientists run simulations forward, but working backwards to find the settings that produced the data is hard because the usual probability formula can't be written down. This review maps tools that learn that shortcut from simulations and finds the approach spreading fast across science.

Abstract

Many domains of science have developed complex simulations to describe phenomena of interest. While these simulations provide high-fidelity models, they are poorly suited for inference and lead to challenging inverse problems. We review the rapidly developing field of simulation-based inference and identify the forces giving new momentum to the field. Finally, we describe how the frontier is expanding so that a broad audience can appreciate the profound change these developments may have on science.

Kyle Cranmer, Johann Brehmer, Gilles Louppe
arXiv:1911.01429 · stat.ML, cs.LG, stat.ME · submitted Nov 4, 2019 · updated Apr 2, 2020
abstract · pdf · html · 10 pages, 3 figures, proceedings for the Sackler Colloquia at the US National Academy of Sciences. v2: fixed typos. v3: clarified text, added references