objective_functions

Response type suitable for optimization

Specification

  • Alias: num_objective_functions

  • Arguments: INTEGER

Child Keywords:

Required/Optional

Description of Group

Dakota Keyword

Dakota Keyword Description

Optional

sense

Whether to minimize or maximize each objective function

Optional

primary_scale_types

How to scale each objective function

Optional

primary_scales

Characteristic values to scale each objective function

Optional

weights

Specify weights for each objective function

Optional

nonlinear_inequality_constraints

Group to specify nonlinear inequality constraints

Optional

nonlinear_equality_constraints

Group to specify nonlinear equality constraints

Optional

scalar_objectives

Number of scalar objective functions

Optional

field_objectives

Number of field objective functions

Description

Specifies the number (1 or more) of objective functions \(f_j\) returned to Dakota for use in the general optimization problem formulation:

\[\begin{split}\begin{eqnarray*} \hbox{minimize:} & & f(\mathbf{x}) = \sum_j{w_j f_j} \\ & & \mathbf{x} \in \Re^{n} \\ \hbox{subject to:} & & \mathbf{g}_{L} \leq \mathbf{g(x)} \leq \mathbf{g}_U \\ & & \mathbf{h(x)}=\mathbf{h}_{t} \\ & & \mathbf{a}_{L} \leq \mathbf{A}_i\mathbf{x} \leq \mathbf{a}_U \\ & & \mathbf{A}_{e}\mathbf{x}=\mathbf{a}_{t} \\ & & \mathbf{x}_{L} \leq \mathbf{x} \leq \mathbf{x}_U \end{eqnarray*}\end{split}\]

Unless responses-objective_functions-sense is specified, Dakota will minimize the objective functions.

The keywords responses-objective_functions-nonlinear_inequality_constraints and responses-objective_functions-nonlinear_equality_constraints specify the number of nonlinear inequality constraints g, and nonlinear equality constraints h, respectively. When interfacing to external applications, the responses must be returned to Dakota in this order in the interface-analysis_drivers-fork-results_file :

  1. objective functions

  2. nonlinear_inequality_constraints

  3. nonlinear_equality_constraints

An optimization problem’s linear constraints are provided to the solver at startup only and do not need to be included in the data returned on every function evaluation. Linear constraints are therefore specified in the variables block through the variables-linear_inequality_constraint_matrix \(A_i\) and variables-linear_equality_constraint_matrix \(A_e\) .

Lower and upper bounds on the design variables x are also specified in the variables block.

The optional keywords relate to scaling the objective functions (for better numerical results), formulating the problem as minimization or maximization, and dealing with multiple objective functions through responses-objective_functions-weights w. If scaling is used, it is applied before multi-objective weighted sums are formed, so, e.g, when both weighting and characteristic value scaling are present the ultimate objective function would be:

\[f = \sum_{j=1}^{n} w_{j} \frac{ f_{j} }{ s_j }\]