ML MCONF NEW: Difference between revisions

From VASP Wiki
No edit summary
 
(14 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{DISPLAYTITLE:ML_MCONF_NEW}}
{{TAGDEF|ML_MCONF_NEW|[integer]|5}}
{{TAGDEF|ML_MCONF_NEW|[integer]|5}}


Description: This tag sets the number of configurations that are stored temporarily as candidates for the training data in the machine learning force field method.
Description: This tag sets the number of configurations that are stored temporarily as candidates for the training data in the machine learning force field method.
----
----
{{NB|warning|This value is empirically set and should usually not be touched.}}
{{NB|warning|This value is close to optimal for on-the-fly learning,  and should usually not be changed. }}


Usually one can employ that the force field doesn't necessary needs to be retrained immediately at every step when a training structure with corresponding local configurations is added. Instead one can also collect candidates and do the learning in a later step for all structures simultaneously. This way saving significant computational cost is saved. Of course learning after every new configurations or after every blocks can have different results, but with not too large block size the difference should be small.  
The use of this tag in combination with the learning algorithms is described here: [[Machine learning force field calculations: Basics#Sampling of training data and local reference configurations|here]].
If force fields are reparameterized ({{TAGO|ML_MODE|select}}), calculations are usually more efficient if this parameter is increased to values around 10-16 and setting {{TAGO|ML_CDOUB|4}}. This is particularly relevant if the ML_AB file is large.


The tag {{TAG|ML_MCONF_NEW}} sets the block size for learning. If the Bayesian error of the force for any atom is above the threshold {{TAG|ML_CTIFOR}} but below {{TAG|ML_CDOUB}}<math>\times</math>{{TAG|ML_CTIFOR}}, the structure is added to the list of new training structures. Whenever the number of candidates is equal to {{TAG|ML_MCONF_NEW}} the new training structures are added to the training structures and the force field is updated. To avoid sampling of too similar structures the next step from which on training structures are allowed to be taken as candidates is set by {{TAG|ML_NMDINT}}. All ab initio calculations within this distance are skipped if the Bayesian error for the force on all atoms is below {{TAG|ML_CDOUB}}<math>\times</math>{{TAG|ML_CTIFOR}}. If the error at any time is above {{TAG|ML_CDOUB}}<math>\times</math>{{TAG|ML_CTIFOR}} immediately the candidates are added to the list of training structure and the force field is updated. This is like an emergency break which won't allow the force field to drift too far away from the ab initio trajectories.
== Related tags and articles ==
 
 
== Related Tags and Sections ==


{{sc|ML_MCONF_NEW|Examples|Examples that use this tag}}
{{sc|ML_MCONF_NEW|Examples|Examples that use this tag}}
----
----
{{TAG|ML_LMLFF}}, {{TAG|ML_MCONF}},{{TAG|ML_CTIFOR}}, {{TAG|ML_CDOUB}}
{{TAG|ML_LMLFF}}, {{TAG|ML_MCONF}}, {{TAG|ML_CTIFOR}}, {{TAG|ML_CDOUB}}


[[Category:INCAR]][[Category:Machine Learning]][[Category:Machine Learned Force Fields]][[Category: Alpha]]
[[Category:INCAR tag]][[Category:Machine-learned force fields]]

Latest revision as of 15:32, 19 October 2023

ML_MCONF_NEW = [integer]
Default: ML_MCONF_NEW = 5 

Description: This tag sets the number of configurations that are stored temporarily as candidates for the training data in the machine learning force field method.


Warning: This value is close to optimal for on-the-fly learning, and should usually not be changed.

The use of this tag in combination with the learning algorithms is described here: here. If force fields are reparameterized (ML_MODE = select), calculations are usually more efficient if this parameter is increased to values around 10-16 and setting ML_CDOUB = 4. This is particularly relevant if the ML_AB file is large.

Related tags and articles

Examples that use this tag


ML_LMLFF, ML_MCONF, ML_CTIFOR, ML_CDOUB