ML ICRITERIA: Difference between revisions

From VASP Wiki
No edit summary
No edit summary
Line 4: Line 4:
----
----


Generally it is recommended to automatically update the criteria {{TAG|ML_FF_CTIFOR}} during machine learning. Details on how and when the update is performed are controlled by {{TAG|ML_FF_CSLOPE}}, {{TAG|ML_FF_CSIG}} and {{TAG|ML_FF_MHIS}}.
Generally it is recommended to automatically update the threshold {{TAG|ML_FF_CTIFOR}} during machine learning. Details on how and when the update is performed are controlled by {{TAG|ML_FF_CSLOPE}}, {{TAG|ML_FF_CSIG}} and {{TAG|ML_FF_MHIS}}.


{{TAG|ML_FF_CTIFOR}} is generally set to the average of the  Bayesian errors of the forces stored in a history. The number of entries in the history are controlled by  {{TAG|ML_FF_MHIS}}. To avoid that noisy data or an abrupt jump of the Bayesian error causes issues, the standard error of the history must be below the threshold  {{TAG|ML_FF_CSIG}}, for the update to take place. Furthermore, the slope of the stored data must be below the threshold  {{TAG|ML_FF_CSLOPE}} (we recommend to set only  {{TAG|ML_FF_CSIG}}).
{{TAG|ML_FF_CTIFOR}} is generally set to the average of the  Bayesian errors of the forces stored in a history. The number of entries in the history are controlled by  {{TAG|ML_FF_MHIS}}. To avoid that noisy data or an abrupt jump of the Bayesian error causes issues, the standard error of the history must be below the threshold  {{TAG|ML_FF_CSIG}}, for the update to take place. Furthermore, the slope of the stored data must be below the threshold  {{TAG|ML_FF_CSLOPE}} (we recommend to set only  {{TAG|ML_FF_CSIG}}).


If the previous conditions are met, the criteria {{TAG|ML_FF_CTIFOR}} is updated. To avoid too abrupt changes the average Bayesian error can be mixed with the current value of  {{TAG|ML_FF_CTIFOR}}. The mixing ratio can be determined by the tag {{TAG|ML_FF_XMIX}} (default is no mixing).
If the previous conditions are met, the threshold {{TAG|ML_FF_CTIFOR}} is updated. To avoid too abrupt changes the average Bayesian error can be mixed with the current value of  {{TAG|ML_FF_CTIFOR}}. The mixing ratio can be determined by the tag {{TAG|ML_FF_XMIX}} (default is no mixing).


== Related Tags and Sections ==
== Related Tags and Sections ==

Revision as of 09:41, 12 June 2021

ML_FF_LCRITERIA = [logical]
Default: ML_FF_LCRITERIA = .TRUE. 

Description: Decides whether the threshold (ML_FF_CTIFOR) is updated in the machine learning force field methods. ML_FF_CTIFOR determines whether a first principles calculations is performed.


Generally it is recommended to automatically update the threshold ML_FF_CTIFOR during machine learning. Details on how and when the update is performed are controlled by ML_FF_CSLOPE, ML_FF_CSIG and ML_FF_MHIS.

ML_FF_CTIFOR is generally set to the average of the Bayesian errors of the forces stored in a history. The number of entries in the history are controlled by ML_FF_MHIS. To avoid that noisy data or an abrupt jump of the Bayesian error causes issues, the standard error of the history must be below the threshold ML_FF_CSIG, for the update to take place. Furthermore, the slope of the stored data must be below the threshold ML_FF_CSLOPE (we recommend to set only ML_FF_CSIG).

If the previous conditions are met, the threshold ML_FF_CTIFOR is updated. To avoid too abrupt changes the average Bayesian error can be mixed with the current value of ML_FF_CTIFOR. The mixing ratio can be determined by the tag ML_FF_XMIX (default is no mixing).

Related Tags and Sections

ML_FF_LMLFF, ML_FF_CTIFOR, ML_FF_CSLOPE, ML_FF_CSIG, ML_FF_MHIS, ML_FF_XMIX

Examples that use this tag