allocation_control

Sample allocation approach for multifidelity expansions

Specification

  • Alias: None

  • Arguments: None

Child Keywords:

Required/Optional

Description of Group

Dakota Keyword

Dakota Keyword Description

Required (Choose One)

Multifidelity Sample Allocation Control

greedy

Sample allocation based on greedy refinement within multifidelity function train

Description

Multifidelity surrogate approaches, including polynomial chaos, stochastic collocation, and function train, can optionally employ a integrated greedy competition across the model sequence, where each model index can supply one or more refinement candidates which are competed to determine the candidate with the greatest impact on the QoI statistics per unit cost. This greedy competition implicitly determines the optimal sample allocation across model indices.

Default Behavior

The default, when allocation_control is not specified, is to compute or adapt separately for each model index (individual instead of integrated refinement).