A Bayesian machine scientist to aid in the solution of challenging scientific problems

• DOI: 10.1126/sciadv.aav6971
• Publication Date: 2020-01-31T22:47:47Z
• Authors: Roger Guimerà; Ignasi Reichardt; Antoni Aguilar-Mogas; Francesco A. Mass...; Manuel Miranda; Jordi Pallarès; Marta Sales-Pardo
• Keywords: FOS: Computer and information sciences; 0301 basic medicine; Computer Science - Machine Learning; 0303 health sciences; FOS: Physical sciences; Machine Learning (stat.ML); 310; 7. Clean energy; Machine Learning (cs.LG); 03 medical and health sciences; Statistics - Machine Learning; 13. Climate action; Physics - Data Analysis, Statistics and Probability; Research Articles; Data Analysis, Statistics and Probability (physics.data-an)

Closed-form, interpretable mathematical models have been instrumental for advancing our understanding of the world; with data revolution, we may now be in a position to uncover new such many systems from physics social sciences. However, deal increasing amounts data, need "machine scientists" that are able extract these automatically data. Here, introduce Bayesian machine scientist, which establishes plausibility using explicit approximations exact marginal posterior over and its prior expectations about by learning large empirical corpus expressions. It explores space Markov chain Monte Carlo. We show this approach uncovers accurate synthetic real provides out-of-sample predictions more than those existing approaches other nonparametric methods.