experimental_gaussian_process

Use the Gaussian process regression surrogate from the surrogates module

Specification

• Alias: None

• Arguments: None

Child Keywords:

Required/Optional	Description of Group	Dakota Keyword	Dakota Keyword Description
Optional		trend	This keyword enables the use of deterministic polynomial trend function
Optional		num_restarts	Number of optimization restarts for L-BFGS-B
Optional (Choose One)	Nugget	nugget	Value for the fixed nugget parameter
		find_nugget	Use regression to estimate the nugget.
Optional		options_file	Filename for a YAML file that specifies Gaussian process options
Optional		export_approx_variance_file	Output file for surrogate model variance evaluations
Optional		export_model	Exports surrogate model in user-specified format(s)
Optional		import_model	Import surrogate model from archive file

Description

This Gaussian process implementation is contained in Dakota’s surrogates module and is considered experimental. It uses gradient-based optimization with restarts to determine hyperparmeters and trend coefficients. Nugget and trend estimation are optional.