class_bias_ensemble

probly.transformation.class_bias_ensemble(base: Predictor[In, Out], num_members: int, reset_params: bool = True, tobias_value: int = 100) → ClassBiasEnsemblePredictor[In, Out][source]

Create an ensemble with class-specific final-layer bias initialization based on [LohrHMHullermeier25].

Parameters:

  • base – The base model to be used for the class-bias ensemble.

  • num_members – The number of members in the class-bias ensemble.

  • reset_params – Whether to reset the parameters of each member.

  • tobias_value – The value to use for the class bias initialization.

Returns:

The class-bias ensemble predictor.