autoprognosis.explorers.core.selector module

class PipelineSelector(classifier: str, calibration: List[int] = [0, 1, 2], imputers: List[str] = [], feature_scaling: List[str] = ['feature_normalizer', 'uniform_transform', 'normal_transform', 'maxabs_scaler', 'minmax_scaler', 'scaler', 'nop'], feature_selection: List[str] = ['nop', 'pca', 'fast_ica'], classifier_category: str = 'classifier')

Bases: object

AutoML wrapper for pipelines

Parameters:

• classifier – str Last estimator of the pipeline, the final classifier.

• calibration – int Type of calibration to use. 0 - none, 1 - sigmoid, 2 - isotonic.

• imputers – list list of imputers to sample from.

• feature_scaling – list list of feature scaling transformers to sample from.

• feature_selection – list list of feature selection methods ti sample from

• classifier_category – str task type: “classifier” or “risk_estimation”

get_pipeline(search_domains: List[Params], hyperparams: List) → PipelineMeta

get_pipeline_from_named_args(**kwargs: Any) → PipelineMeta

get_pipeline_from_template(model_list: List, args: Dict) → PipelineMeta

get_pipeline_template(search_domains: List[Params], hyperparams: List) → Tuple[List, Dict]

hyperparameter_space() → List

name() → str

sample_hyperparameters(trial: optuna.trial.Trial) → Dict

sample_hyperparameters_np() → Dict