ML CDOUB: Difference between revisions
No edit summary |
No edit summary |
||
| Line 4: | Line 4: | ||
Description: This tag controls the criterion for "enforced" DFT calculations within the machine learning force field method. | Description: This tag controls the criterion for "enforced" DFT calculations within the machine learning force field method. | ||
---- | ---- | ||
The usage of this tag in combination with the learning algorithms is described here: [[Machine learning force field calculations: Important algorithms#Sampling of training data and local reference configurations|here]]. | |||
If at any time, the estimated force errors are {{TAG|ML_CDOUB}} times larger than the Bayesian threshold (i.e. "critically" high), a first principles calculation is performed and a new force field is immediately generated (even if the counter for sampling is below the minimum amount of sampled structures {{TAG|ML_NMDINT}}). | If at any time, the estimated force errors are {{TAG|ML_CDOUB}} times larger than the Bayesian threshold (i.e. "critically" high), a first principles calculation is performed and a new force field is immediately generated (even if the counter for sampling is below the minimum amount of sampled structures {{TAG|ML_NMDINT}}). | ||
Revision as of 13:18, 21 October 2021
ML_CDOUB = [real]
Default: ML_CDOUB = 2.0
Description: This tag controls the criterion for "enforced" DFT calculations within the machine learning force field method.
The usage of this tag in combination with the learning algorithms is described here: here.
If at any time, the estimated force errors are ML_CDOUB times larger than the Bayesian threshold (i.e. "critically" high), a first principles calculation is performed and a new force field is immediately generated (even if the counter for sampling is below the minimum amount of sampled structures ML_NMDINT).
Related Tags and Sections
ML_LMLFF, ML_CTIFOR, ML_NMDINT