If optGPEveryN > m, that is, the GP hyperparameter optimization cadence is greater than the number of new points to find per iteration, the GP hyperparameters will never be re-optimized. approxposterior should at least warn the user if this is occurring. See line 681 in approx.py for where this can be fixed and/or where an error/warning could be raised.
dflemin3
commented
4 years ago
If optGPEveryN > m, that is, the GP hyperparameter optimization cadence is greater than the number of new points to find per iteration, the GP hyperparameters will never be re-optimized. approxposterior should at least warn the user if this is occurring. See line 681 in approx.py for where this can be fixed and/or where an error/warning could be raised.