Perform bounds-scaling on response values prior to surrogate emulation


  • Alias: None

  • Arguments: None

  • Default: use original data


Particularly in multifidelity methods based on regression solutions at each level of a model hierarchy, scaling can be an issue since the magnitude of discrepancy data may decay rapidly and regression solvers may employ absolute tolerances in places.

By activating response_scaling, the set of response data used for each recovery is scaled to [0,1] based on the minimum value and range of the data set. This prevents loss of accuracy due to imbalances in scale across model levels.