NestedTabularMLAlgo

class lightautoml.pipelines.ml.nested_ml_pipe.NestedTabularMLAlgo(ml_algo, tuner=None, refit_tuner=False, cv=5, n_folds=None)[source]

Bases: lightautoml.ml_algo.base.TabularMLAlgo, lightautoml.pipelines.selection.base.ImportanceEstimator

Wrapper for MLAlgo to make it trainable over nested folds. Limitations - only for TabularMLAlgo.

property params

Parameters of ml_algo.

Return type: dict

init_params_on_input(train_valid_iterator)[source]

Init params depending on input data.

Return type: dict
Returns: dict with model hyperparameters.

fit_predict_single_fold(train, valid)[source]

Implements training and prediction on single fold.

Parameters

train (Union[NumpyDataset, PandasDataset]) – TabularDataset to train.
valid (Union[NumpyDataset, PandasDataset]) – TabularDataset to validate.

Return type

Tuple[Any, ndarray]

Returns

Tuple (model, predicted_values).

fit(train_valid)[source]

Just to be compatible with ImportanceEstimator.

Parameters: train_valid (TrainValidIterator) – Classic cv iterator.