![]() |
Dakota Reference Manual
Version 6.15
Explore and Predict with Confidence
|
Stopping criterion based on number of refinement iterations within the multilevel sample allocation
Alias: none
Argument(s): INTEGER
Default: 100 (exceptions: fsu_cvt , local_reliability: 25; global_{reliability , interval_est , evidence} / efficient_global: 25*n)
Multilevel sampling is an iterative procedure that estimates the optimal number of samples for each level based on cost and observed variance. On each iteration, additional samples are performed and more accurate variance estimates are computed, leading to updated sample allocations. The process terminates when either no additional samples are allocated or the max_iterations control is enforced.
Default Behavior
The default value for max_iterations varies by method. For multilevel_sampling, the default value is 25.