estimator_performance

Return estimator performance as the final results of a UQ method

Specification

  • Alias: None

  • Arguments: None

Description

Instead of QoI moments and level mappings, return the performance of the estimator in terms of its variance and aggregate cost. These metrics are targeted in the context of model tuning for multifidelity UQ methods.

Usage Tips Estimator variance can be minimized when total budget is specified. Alternatively, total budget can be minimized when estimator accuracy is specified. This is managed by the nested model response mappings that define the outer loop objective and constraints, sharing conceptual similarity to the RIA and PMA formulations for optimization with local reliability (refer to RBDO methods in the Theory Manual).

Examples

method,
 id_method = 'UQ'
 model_pointer = 'ENSEMBLE'
 multifidelity_sampling
   max_function_evaluations = 1000
   pilot_samples = 25
   seed = 8674132
   solution_mode pilot_projection
   final_statistics estimator_performance