$extrastylesheet
Dakota
Version 6.2
|
Derived approximation class for Surfpack approximation classes. Interface between Surfpack and Dakota. More...
Public Member Functions | |
SharedSurfpackApproxData () | |
default constructor | |
SharedSurfpackApproxData (const String &approx_type, const UShortArray &approx_order, size_t num_vars, short data_order, short output_level) | |
alternate constructor | |
SharedSurfpackApproxData (ProblemDescDB &problem_db, size_t num_vars) | |
standard constructor: Surfpack surface of appropriate type will be created | |
~SharedSurfpackApproxData () | |
destructor | |
Private Member Functions | |
void | add_sd_to_surfdata (const Pecos::SurrogateDataVars &sdv, const Pecos::SurrogateDataResp &sdr, short fail_code, SurfData &surf_data) |
add Pecos::SurrogateData::SurrogateData{Vars,Resp} to SurfData, accounting for buildDataOrder available | |
void | copy_matrix (const RealSymMatrix &rsm, SurfpackMatrix< Real > &surfpack_matrix) |
copy RealSymMatrix to SurfpackMatrix (Real type only) | |
void | merge_variable_arrays (const RealVector &cv, const IntVector &div, const RealVector &drv, RealArray &ra) |
merge cv, div, and drv vectors into a single ra array | |
void | sdv_to_realarray (const Pecos::SurrogateDataVars &sdv, RealArray &ra) |
aggregate {continuous,discrete int,discrete real} variables from SurrogateDataVars into ra | |
void | vars_to_realarray (const Variables &vars, RealArray &ra) |
aggregate {active,all} {continuous,discrete int,discrete real} variables into ra | |
Private Attributes | |
unsigned short | approxOrder |
order of polynomial approximation | |
String | exportModelName |
A Surfpack model name for saving the surrogate model. | |
StringArray | diagnosticSet |
set of diagnostic metrics | |
bool | crossValidateFlag |
whether to perform cross validation | |
unsigned | numFolds |
number of folds for CV | |
Real | percentFold |
percentage of data for CV | |
bool | pressFlag |
whether to perform PRESS | |
Friends | |
class | SurfpackApproximation |
class | VPSApproximation |
Derived approximation class for Surfpack approximation classes. Interface between Surfpack and Dakota.
The SharedSurfpackApproxData class is the interface between Dakota and Surfpack. Based on the information in the ProblemDescDB that is passed in through the constructor, SharedSurfpackApproxData builds a Surfpack Surface object that corresponds to one of the following data-fitting techniques: polynomial regression, kriging, artificial neural networks, radial basis function network, or multivariate adaptaive regression splines (MARS).
SharedSurfpackApproxData | ( | const String & | approx_type, |
const UShortArray & | approx_order, | ||
size_t | num_vars, | ||
short | data_order, | ||
short | output_level | ||
) |
alternate constructor
On-the-fly constructor which uses mostly Surfpack model defaults.
References Dakota::abort_handler(), SharedSurfpackApproxData::approxOrder, and SharedApproxData::approxType.
SharedSurfpackApproxData | ( | ProblemDescDB & | problem_db, |
size_t | num_vars | ||
) |
standard constructor: Surfpack surface of appropriate type will be created
Initialize the embedded Surfpack surface object and configure it using the specifications from the input file. Data for the surface is created later.
References SharedSurfpackApproxData::approxOrder, SharedApproxData::approxType, ProblemDescDB::get_short(), and ProblemDescDB::get_string().