ML W1: Difference between revisions

Revision as of 20:11, 11 October 2021

ML_W1 = [real]
Default: ML_W1 = 0.1

Description: This tag defines the weight $\beta$ for the radial (and angular) descriptor within the machine learning force field method (see this section).

The weight for the angular descriptor $W_{2}$ is internally computed from the weight of the radial descriptor $W_{1}$ as:

$W_{2}=1.0-W_{1}.$

The value for ML_W1 must be chosen in the interval $[0,1]$ .

By default, the angular and radial descriptors are both used although the latter is weighed less. In principle a weight of 0 for one of them is selectable which allows the code to internally skip the respective computation. However, it is generally recommended to use both descriptors to achieve satisfying training results.

Related Tags and Sections

ML_LMLFF, ML_RCUT1, ML_RCUT2, ML_SION1, ML_SION2

Examples that use this tag

@@ Line 1: / Line 1: @@
-{{TAGDEF|ML_W1|[real]|0.5}}
+{{TAGDEF|ML_W1|[real]|0.1}}
-Description: This tag defines the weight for the radial (and angular) descriptor within the machine learning force field method.
+Description: This tag defines the weight <math>\beta</math> for the radial (and angular) descriptor within the machine learning force field method (see [[Machine learning force field: Theory#Potential energy fitting|this section]]).
 ----
-The weight for the angular descriptor <math>W_{2}</math> is internally derived from the weight of the radial descriptor <math>W_{1}</math> as:
+The weight for the angular descriptor <math>W_{2}</math> is internally computed from the weight of the radial descriptor <math>W_{1}</math> as:
 <math>W_{2}=1.0-W_{1}.</math>
-By default this descriptor is used together with the angular descriptor with equal weights. If the weight of either becomes 0 then the internal routines for that descriptor are mostly skipped.
+The value for {{TAG|ML_W1}} must be chosen in the interval <math>[0, 1]</math>.
+By default, the angular and radial descriptors are both used although the latter is weighed less. In principle a weight of 0 for one of them is selectable which allows the code to internally skip the respective computation. However, it is generally recommended to use both descriptors to achieve satisfying training results.
 == Related Tags and Sections ==