Hello, I’ve been using Openturns for the last two-three weeks to develop some Kriging models for my thesis. I was wondering if it is possible to put in hyperparamaters (that are optimized) into an “UserDefinedCovarianceModel”?
I would like to play around with models other then the ones given by OT itself. If this is possible, is there an example that I could follow somewhere?
Hi @smollie37 ! Unfortunately, UserDefinedCovarianceModel admits no parameter. If the covariance kernel you want to implement is stationary, then perhaps the class StationaryFunctionalCovarianceModel could help you: it would allow you to optimize the scale and amplitude hyperparameters of an arbitrary stationary covariance kernel.
@josephmure, thank you for the quick response. I’ll have a look on it. Perhaps this will get me what I need. 