ML IREG: Difference between revisions
No edit summary |
No edit summary |
||
Line 1: | Line 1: | ||
{{TAGDEF|ML_FF_IREG_MB|[integer]|2}} | {{TAGDEF|ML_FF_IREG_MB|[integer]|2}} | ||
Description: This tag specifies whether the | Description: This tag specifies whether the regularization parameters are kept constant or not in the machine learning force field method. | ||
---- | ---- | ||
The following cases are possible for this tag: | The following cases are possible for this tag: |
Revision as of 10:09, 6 June 2019
ML_FF_IREG_MB = [integer]
Default: ML_FF_IREG_MB = 2
Description: This tag specifies whether the regularization parameters are kept constant or not in the machine learning force field method.
The following cases are possible for this tag:
- ML_FF_IREG_MB=1: The (initial) precision (ML_FF_SIGV0_MB) and noise (ML_FF_SIGW0_MB) parameters are kept constant.
- ML_FF_IREG_MB=2: The parameters are optimized (default).
For the optimization of the noise parameter see following section: On-the-fly machine learning force field generation using Bayesian linear regression.
Related Tags and Sections
ML_FF_LMLFF, ML_FF_SIGV0_MB, ML_FF_SIGW0_MB