approx_subproblem

Identify functions to be included in surrogate merit function

Specification

  • Alias: None

  • Arguments: None

  • Default: original_primary original_constraints

Child Keywords:

Required/Optional

Description of Group

Dakota Keyword

Dakota Keyword Description

Required (Choose One)

Objective Formulation

original_primary

Construct approximations of all primary functions

single_objective

Construct approximation a single objective functions only

augmented_lagrangian_objective

Augmented Lagrangian approximate subproblem formulation

lagrangian_objective

Lagrangian approximate subproblem formulation

Required (Choose One)

Constraint Formulation

original_constraints

Use the constraints directly

linearized_constraints

Use linearized approximations to the constraints

no_constraints

Don’t use constraints

Description

First, the “primary” functions (that is, the objective functions or calibration terms) in the approximate subproblem can be selected to be surrogates of the original primary functions ( original_primary), a single objective function ( single_objective) formed from the primary function surrogates, or either an augmented Lagrangian merit function ( augmented_lagrangian_objective) or a Lagrangian merit function ( lagrangian_objective) formed from the primary and secondary function surrogates. The former option may imply the use of a nonlinear least squares method, a multiobjective optimization method, or a single objective optimization method to solve the approximate subproblem, depending on the definition of the primary functions. The latter three options all imply the use of a single objective optimization method regardless of primary function definition. Second, the surrogate constraints in the approximate subproblem can be selected to be surrogates of the original constraints ( original_constraints) or linearized approximations to the surrogate constraints ( linearized_constraints), or constraints can be omitted from the subproblem ( no_constraints).