https://vasp.at/wiki/index.php?action=history&feed=atom&title=ML_CALGOML CALGO - Revision history2026-04-15T17:37:32ZRevision history for this page on the wikiMediaWiki 1.43.8https://vasp.at/wiki/index.php?title=ML_CALGO&diff=27674&oldid=prevKarsai: Created page with "{{DISPLAYTITLE:ML_CALGO}} {{TAGDEF|ML_CALGO|[integer]|0}} Description: Chooses error estimation type for on-the-fly training or reselection of local referenc configurations. ---- This tag chooes which algorithm is employed for the error estimation in {{TAG|ML_MODE}}=''TRAIN'' or ''SELECT''. The following two choices are available: *{{TAG|ML_CALGO}}=0: Bayesian error estimation. Constant or variable threshold. Default. *{{TAG|ML_CALGO}}=1: Spilling factor. Constant thr..."2024-12-18T14:39:33Z<p>Created page with "{{DISPLAYTITLE:ML_CALGO}} {{TAGDEF|ML_CALGO|[integer]|0}} Description: Chooses error estimation type for on-the-fly training or reselection of local referenc configurations. ---- This tag chooes which algorithm is employed for the error estimation in {{TAG|ML_MODE}}=''TRAIN'' or ''SELECT''. The following two choices are available: *{{TAG|ML_CALGO}}=0: Bayesian error estimation. Constant or variable threshold. Default. *{{TAG|ML_CALGO}}=1: Spilling factor. Constant thr..."</p>
<p><b>New page</b></p><div>{{DISPLAYTITLE:ML_CALGO}}<br />
{{TAGDEF|ML_CALGO|[integer]|0}}<br />
<br />
Description: Chooses error estimation type for on-the-fly training or reselection of local referenc configurations.<br />
----<br />
<br />
This tag chooes which algorithm is employed for the error estimation in {{TAG|ML_MODE}}=''TRAIN'' or ''SELECT''. The following two choices are available:<br />
*{{TAG|ML_CALGO}}=0: Bayesian error estimation. Constant or variable threshold. Default. <br />
*{{TAG|ML_CALGO}}=1: Spilling factor. Constant threhold. <br />
<br />
In both modes an ab-initio calculation is carried out if the value of the error estimate is above a threshold specified by {{TAG|ML_CTIFOR}}. In both algorithms the estimators have different units, values and hence defaults for this threshold. <br />
In contrast to the Bayesian error estimation which can be run in many different modes for the threhold update (see {{TAG|ML_ICRITERIA}}), the spilling factor can only be used with a constant threshold ({{TAG|ML_ICRITERIA}}=0).<br />
<br />
== Related tags and articles ==<br />
{{TAG|ML_LMLFF}}, {{TAG|ML_MODE}}, {{TAG|ML_ICRITERIA}}, {{TAG|ML_CTIFOR}}<br />
----<br />
<br />
[[Category:INCAR tag]][[Category:Machine-learned force fields]]</div>Karsai