rcond_tolerance

Throttle multilevel BLUE to only solve for allocations using groups with prescribed conditioning in their group covariances

Specification

  • Alias: None

  • Arguments: None

  • JSON argument key: None

Description

The number of groups in ML BLUE is initially defined using all combinations of the model indices, and group covariances are computed for every group. This throttle then retains only the groups for which the group covariances have an inverse condition number (“rcond”) greater than the provided tolerance. This is closely related to best_conditioned except that the number of retained groups is not known a priori.

Examples

method
  model_pointer = 'ENSEMBLE'
  multilevel_blue
    pilot_samples = 50 seed = 8674132
    group_throttle rcond_tolerance = 1.e-12
    max_function_evaluations = 1000
{
  "method": [
    {
      "multilevel_blue": {
        "model_pointer": "ENSEMBLE",
        "pilot_samples": {"counts": [50]},
        "seed_sequence": [8674132],
        "group_throttle": {"rcond_tolerance": 1e-12},
        "max_function_evaluations": 1000
      }
    }
  ]
}