ML CSLOPE: Difference between revisions

From VASP Wiki
No edit summary
mNo edit summary
(9 intermediate revisions by 2 users not shown)
Line 1: Line 1:
{{TAGDEF|ML_FF_CSLOPE|[real]|<math>0.1</math>}}
{{TAGDEF|ML_FF_CSLOPE|[real]|<math>0.2</math>}}


Description: Parameter used in the automatic determination of threshold for Bayesian error estimation in the machine learning force field method.
Description: Parameter used in the automatic determination of threshold for Bayesian error estimation in the machine learning force field method.
----
----
 
----For details please read entry {{TAG|ML_FF_LCRITERIA}} first. The parameter {{TAG|ML_FF_CTIFOR}} is only updated, if the absolute of the slope of the collected Bayesian errors is below {{TAG|ML_FF_CSLOPE}} times the mean of the collected Bayesian errors. In practice, the slope and the standard errors are correlated: typically the standard error is at least twice the slope. We recommend to vary only {{TAG|ML_FF_CSIG}} and keep {{TAG|ML_FF_CSLOPE}} fixed to its default value.
{{TAG|ML_FF_CSLOPE}} is a threshold for slope of the stored estimated errors used to determine the criteria when {{TAG|ML_FF_LCRITERIA}}=''.TRUE.'' is set.
 
== Related Tags and Sections ==
== Related Tags and Sections ==
{{TAG|ML_FF_LMLFF}}, {{TAG|ML_FF_LCRITERIA}}, {{TAG|ML_FF_CSIG}}  
{{TAG|ML_FF_LMLFF}}, {{TAG|ML_FF_IERR}}, {{TAG|ML_FF_ISAMPLE}}, {{TAG|ML_FF_LCRITERIA}}, {{TAG|ML_FF_CSIG}}, {{TAG|ML_FF_MHIS}}  


{{sc|ML_FF_CSLOPE|Examples|Examples that use this tag}}
{{sc|ML_FF_CSLOPE|Examples|Examples that use this tag}}
----
----


[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]][[Category:VASP6]]
[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]][[Category: Alpha]]

Revision as of 08:17, 16 April 2021

ML_FF_CSLOPE = [real]
Default: ML_FF_CSLOPE =  

Description: Parameter used in the automatic determination of threshold for Bayesian error estimation in the machine learning force field method.



For details please read entry ML_FF_LCRITERIA first. The parameter ML_FF_CTIFOR is only updated, if the absolute of the slope of the collected Bayesian errors is below ML_FF_CSLOPE times the mean of the collected Bayesian errors. In practice, the slope and the standard errors are correlated: typically the standard error is at least twice the slope. We recommend to vary only ML_FF_CSIG and keep ML_FF_CSLOPE fixed to its default value.

Related Tags and Sections

ML_FF_LMLFF, ML_FF_IERR, ML_FF_ISAMPLE, ML_FF_LCRITERIA, ML_FF_CSIG, ML_FF_MHIS

Examples that use this tag