ML EPS LOW: Difference between revisions

From VASP Wiki
No edit summary
 
(9 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{TAGDEF|ML_EPS_LOW|[real]|1E-9}} (vasp.6.3.1 default was 1E-10, see comments below)
{{DISPLAYTITLE:ML_EPS_LOW}}
{{TAGDEF|ML_EPS_LOW|[real]}}
{{DEF|ML_EPS_LOW|1E-11|for {{TAG|ML_MODE}} {{=}} SELECT, REFIT|1E-9|else (vasp.6.3.0 default was 1E-10, see comments below)}}


Description: Threshold for the CUR algorithm used in the sparsification of local reference configurations within the machine learning force fields.  
Description: Threshold for the CUR algorithm used in the sparsification of local reference configurations within the machine learning force fields.  
----
----
This value sets the threshold for the eigenvalues that contribute to the leverage scoring used in the CUR algorithm for the rank compression ("sparsification") of the local configurations (for details see appendix E of reference {{cite|jinnouchi2:arx:2019}}). Small eigenvalues and those columns (local configurations) that are strongly connected
This value sets the threshold for the eigenvalues that contribute to the leverage scoring used in the CUR algorithm for the rank compression ("sparsification") of the local configurations (for details see appendix E of reference {{cite|jinnouchi2:arx:2019}}). Small eigenvalues and those columns (local configurations) that are strongly connected
with these small eigenvalues are removed by the sparsification routines. The default value is fairly well balanced, and we do not  recommend to reduce the threshold to values  
with these small eigenvalues are removed by the sparsification routines. The default value is fairly well balanced, and we do not  recommend to increase the threshold to values  
below 1E-9. Using smaller values than 1E-9, does not improve the MLFF if Bayesian regression is used.  
larger than 1E-7. Also using smaller values than 1E-9 does not improve the MLFF if Bayesian regression is used
(but it can be benficial for SVD).  


The description how to choose {{TAG|ML_EPS_LOW}} for accurate force fields using SVD is given [[Machine learning force field calculations: Basics#Accurate force fields|here]].
The description how to choose {{TAG|ML_EPS_LOW}} for accurate force fields is given [[Best practices for machine-learned force fields#Accurate force fields|here]].


On the theory of the sparsification of local reference configurations see
On the theory of the sparsification of local reference configurations see
[[Machine learning force field: Theory#Sparsification of local reference configurations|here]].
[[Machine learning force field: Theory#Sparsification of local reference configurations|here]].


== Related Tags and Sections ==
== Related tags and articles ==
{{TAG|ML_LMLFF}}, {{TAG|ML_MB}}, {{TAG|ML_EPS_REG}}
{{TAG|ML_LMLFF}}, {{TAG|ML_MB}}, {{TAG|ML_EPS_REG}}, {{TAG|ML_IALGO_LINREG}}


{{sc|ML_EPS_LOW|Examples|Examples that use this tag}}
{{sc|ML_EPS_LOW|Examples|Examples that use this tag}}
Line 23: Line 26:
----
----


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

Latest revision as of 11:42, 15 May 2023

ML_EPS_LOW = [real] 

Default: ML_EPS_LOW = 1E-11 for ML_MODE = SELECT, REFIT
= 1E-9 else (vasp.6.3.0 default was 1E-10, see comments below)

Description: Threshold for the CUR algorithm used in the sparsification of local reference configurations within the machine learning force fields.


This value sets the threshold for the eigenvalues that contribute to the leverage scoring used in the CUR algorithm for the rank compression ("sparsification") of the local configurations (for details see appendix E of reference [1]). Small eigenvalues and those columns (local configurations) that are strongly connected with these small eigenvalues are removed by the sparsification routines. The default value is fairly well balanced, and we do not recommend to increase the threshold to values larger than 1E-7. Also using smaller values than 1E-9 does not improve the MLFF if Bayesian regression is used (but it can be benficial for SVD).

The description how to choose ML_EPS_LOW for accurate force fields is given here.

On the theory of the sparsification of local reference configurations see here.

Related tags and articles

ML_LMLFF, ML_MB, ML_EPS_REG, ML_IALGO_LINREG

Examples that use this tag

References