Action: OPT_ADAM
| Module | ves |
|---|---|
| Description | Usage |
| Adaptive moment estimation (ADAM) optimizer. |   |
| output value | type |
| a scalar | scalar |
Details and examples
Adaptive moment estimation (ADAM) optimizer.
This optimizer is still experimental and not fully documented. The syntax might change. Restarting does not work. We recommend to use the averaged stochastic gradient decent optimizer (OPT_AVERAGED_SGD) for now.
Output components
This action can calculate the values in the following table when the associated keyword is included in the input for the action. These values can be referenced elsewhere in the input by using this Action's label followed by a dot and the name of the value required from the list below.
| Name | Type | Keyword | Description |
|---|---|---|---|
| gradrms | scalar | MONITOR_INSTANTANEOUS_GRADIENT | the root mean square value of the coefficient gradient |
| gradmax | scalar | MONITOR_INSTANTANEOUS_GRADIENT | the largest absolute value of the coefficient gradient |
Full list of keywords
The following table describes the keywords and options that can be used with this action
| Keyword | Type | Default | Description |
|---|---|---|---|
| BIASThis keyword do not have examples | compulsory | none | the label of the VES bias to be optimized |
| STRIDEThis keyword do not have examples | compulsory | none | the frequency of updating the coefficients given in the number of MD steps |
| COEFFS_FILEThis keyword do not have examples | compulsory | coeffs.data | the name of output file for the coefficients |
| COEFFS_OUTPUTThis keyword do not have examples | compulsory | 100 | how often the coefficients should be written to file |
| STEPSIZEThis keyword do not have examples | compulsory | none | the step size used for the optimization |
| COEFFS_FMTThis keyword do not have examples | optional | not used | specify format for coefficient file(s) (useful for decrease the number of digits in regtests) |
| COEFFS_SET_ID_PREFIXThis keyword do not have examples | optional | not used | suffix to add to the filename given in FILE to identify the bias, should only be given if a single filename is given in FILE when optimizing multiple biases |
| INITIAL_COEFFSThis keyword do not have examples | optional | not used | the name(s) of file(s) with the initial coefficients |
| MONITOR_INSTANTANEOUS_GRADIENTThis keyword do not have examples | optional | false | if quantities related to the instantaneous gradient should be outputted |
| TARGETDIST_AVERAGES_FILEThis keyword do not have examples | optional | not used | the name of output file for the target distribution averages |
| TARGETDIST_AVERAGES_OUTPUTThis keyword do not have examples | optional | not used | how often the target distribution averages should be written out to file |
| BIAS_OUTPUTThis keyword do not have examples | optional | not used | how often the bias(es) should be written out to file |
| FES_OUTPUTThis keyword do not have examples | optional | not used | how often the FES(s) should be written out to file |
| FES_PROJ_OUTPUTThis keyword do not have examples | optional | not used | how often the projections of the FES(s) should be written out to file |
| RESTARTThis keyword do not have examples | optional | not used | allows per-action setting of restart (YES/NO/AUTO) |
| UPDATE_FROMThis keyword do not have examples | optional | not used | Only update this action from this time |
| UPDATE_UNTILThis keyword do not have examples | optional | not used | Only update this action until this time |
| MULTIPLE_WALKERSThis keyword do not have examples | optional | false | if optimization is to be performed using multiple walkers connected via MPI |
| MASK_FILEThis keyword do not have examples | optional | not used | read in a mask file which allows one to employ different step sizes for different coefficients and/or deactivate the optimization of certain coefficients (by putting values of 0 |
| OUTPUT_MASK_FILEThis keyword do not have examples | optional | not used | Name of the file to write out the mask resulting from using the MASK_FILE keyword |
| TARGETDIST_STRIDEThis keyword do not have examples | optional | not used | stride for updating a target distribution that is iteratively updated during the optimization |
| TARGETDIST_OUTPUTThis keyword do not have examples | optional | not used | how often the dynamic target distribution(s) should be written out to file |
| TARGETDIST_PROJ_OUTPUTThis keyword do not have examples | optional | not used | how often the projections of the dynamic target distribution(s) should be written out to file |
| BETA_1This keyword do not have examples | optional | not used | Parameter for the first moment estimate |
| BETA_2This keyword do not have examples | optional | not used | Parameter for the second moment estimate |
| EPSILONThis keyword do not have examples | optional | not used | Small parameter to avoid division by zero |
| ADAMW_WEIGHT_DECAYThis keyword do not have examples | optional | not used | Weight decay parameter for the AdamW variant |
| AMSGRADThis keyword do not have examples | optional | false | Use the AMSGrad variant |