saealib.SurrogatePrediction

class saealib.SurrogatePrediction(value, std=None, label=None, _tell_f=None, metadata=<factory>)[source]

Bases: object

Unified return type for surrogate model predictions.

Parameters:

  • value (ndarray)

  • std (ndarray | None)

  • label (ndarray | None)

  • _tell_f (ndarray | None)

  • metadata (dict)

value

Predicted values. shape: (n_samples, n_obj)

Type:

np.ndarray

std

Predicted standard deviations (uncertainty). shape: (n_samples, n_obj). None if the surrogate does not provide uncertainty estimates (e.g., RBF interpolation).

Type:

np.ndarray or None

label

Predicted class labels. shape: (n_samples,). None unless the surrogate is a classification model.

Type:

np.ndarray or None

tell_f

Values to assign to offspring.f before calling algorithm.tell(). Falls back to mean when no override is set (_tell_f is None). Pass _tell_f=np.full(..., np.nan) to prevent pbest corruption when the surrogate returns non-objective values (e.g., class probabilities, novelty scores).

Type:

np.ndarray

metadata

Implementation-specific additional information (e.g., SHAP values, gradient estimates).

Type:

dict

Methods

__init__

Method Details

SurrogatePrediction.__init__(value, std=None, label=None, _tell_f=None, metadata=<factory>)
Parameters:

  • value (ndarray)

  • std (ndarray | None)

  • label (ndarray | None)

  • _tell_f (ndarray | None)

  • metadata (dict)

Return type:

None