Number of initial model evaluations used in build phase

Specification

Alias: none

Argument(s): INTEGER

Description

The number of build points or training points used in the initial phase of an algorithm or model construction. Typically these are the initial set of data points used to construct (train) a surrogate model (emulator). If the number of build_samples is less than the minimum number of points required to build the surrogate, Dakota will augment the samples to obtain the minimum required.

Examples

Perform GP-based adaptive importance sampling, building the GP with 100 points and then performing 100 approxmiate evaluations to evaluate the probability.

method
  gpais
    build_samples = 100
    samples_on_emulator = 100
    max_iterations = 5
    response_levels = -1.065