Current browse context:
physics.comp-ph
Change to browse by:
References & Citations
Physics > Computational Physics
Title: Exploring the robust extrapolation of high-dimensional machine learning potentials
(Submitted on 20 Dec 2021 (v1), last revised 22 Apr 2022 (this version, v2))
Abstract: We show that, contrary to popular assumptions, predictions from machine learning potentials built upon high-dimensional atom-density representations almost exclusively occur in regions of the representation space which lie outside the convex hull defined by the training set points. We then propose a perspective to rationalize the domain of robust extrapolation and accurate prediction of atomistic machine learning potentials in terms of the probability density induced by training points in the representation space
Submission history
From: Claudio Zeni [view email][v1] Mon, 20 Dec 2021 10:34:02 GMT (320kb,D)
[v2] Fri, 22 Apr 2022 08:33:16 GMT (799kb,D)
Link back to: arXiv, form interface, contact.