MLFF Spilling Factor Decision Algorithm
Posted: Mon Jun 10, 2024 7:58 am
Hi there,
I'm interested in generating a MLFF for an amorphous polymer system and have been struggling with the high number of FP calculations triggered by the Bayesian error in the current on-the-fly implementation (I am using version 6.4.1).
I came across one paper (https://pubs.acs.org/doi/full/10.1021/a ... tt.3c00293) and it seems that using a spilling factor threshold as opposed to the Bayesian Error can result in more efficient learning for amorphous systems.
I was unable to find any information on the spilling factor decision algorithm on the VASP wiki pages and wondered if anyone could offer some advice on how to implement this?
Best Wishes,
Charlotte
I'm interested in generating a MLFF for an amorphous polymer system and have been struggling with the high number of FP calculations triggered by the Bayesian error in the current on-the-fly implementation (I am using version 6.4.1).
I came across one paper (https://pubs.acs.org/doi/full/10.1021/a ... tt.3c00293) and it seems that using a spilling factor threshold as opposed to the Bayesian Error can result in more efficient learning for amorphous systems.
I was unable to find any information on the spilling factor decision algorithm on the VASP wiki pages and wondered if anyone could offer some advice on how to implement this?
Best Wishes,
Charlotte