saealib.KFoldAccuracyEvaluator#

class saealib.KFoldAccuracyEvaluator(metrics=None, n_splits=5)[source]#

Bases: AccuracyEvaluator

Evaluates surrogate accuracy via k-fold cross-validation.

The training data is split into n_splits folds. For each fold, the surrogate is deep-copied and fitted on the remaining folds, then tested on the held-out fold. Metrics are averaged across folds.

Parameters:

  • metrics (list[SurrogateAccuracyMetric] or None) – Metrics to compute.

  • n_splits (int) – Number of folds. Clamped to n_samples if larger.

Methods

__init__

evaluate

Evaluate via k-fold CV.

Method Details

KFoldAccuracyEvaluator.__init__(metrics=None, n_splits=5)[source]#
Parameters:
KFoldAccuracyEvaluator.evaluate(surrogate, train_x, train_y)[source]#

Evaluate via k-fold CV.

Parameters:

  • surrogate (Surrogate)

  • train_x (np.ndarray)

  • train_y (np.ndarray)

Return type:

SurrogateAccuracy