ConditionalSuperLearnerHyperparameters

Properties

base_learners

Hyperparameters[]

meta_learner

Hyperparameters

inner_resampling_config

ResamplerConfig

expand_search_spaces

boolean

Expand a base learner's search space into one library entry per grid combination, rather than tuning it by inner resampling.

ifwtunable

boolean

Inverse Frequency Weighting in classification.

n_iterationstunable

integer≥ 1

Alternations between fitting the oracle and refitting the experts.

loss

string | null

Per-case loss the oracle minimizes. NULL = squared error, which for classification is the Brier score.

one of"squared_error""log_loss"

init

string

Region each expert starts from: the whole training fold, or a random partition of it.

one of"full""random"

min_region_size

integer≥ 1

Fewest cases an expert's region may hold before the expert keeps its previous fit instead of being refitted.

Relationships