Stopping criterion based on maximum function evaluations




  • Alias: None

  • Arguments: INTEGER

  • Default: 1000


For multilevel and multifidelity sampling methods, this specification provides a computational budget, indicating that the mean squared error (MSE) of the estimator is to be minimized within this budget. In these methods, the MSE is typically comprised of known estimator variance (stochastic error), neglecting unknown residual bias (deterministic error for the highest-fidelity/most-resolved model).

The budget is provided in terms of the equivalent number of high-fidelity/fine-resolution evaluations. Default value is SZ_MAX (the maximum storable size_t), which is a special value indicating that a budget has not been specified.