TrainValidIterator

class lightautoml.validation.base.TrainValidIterator(train, **kwargs)[source]

Bases: object

Abstract class to train/validation iteration.

Train/valid iterator: should implement __iter__ and __next__ for using in ml_pipeline.

property features

Dataset features names.

Returns

List of features names.

__init__(train, **kwargs)[source]
Parameters
  • train (~Dataset) – Train dataset.

  • **kwargs – Key-word parameters.

get_validation_data()[source]

Abstract method. Get validation sample.

Return type

LAMLDataset

apply_feature_pipeline(features_pipeline)[source]

Apply features pipeline on train data.

Parameters

features_pipeline (FeaturesPipeline) – Composite transformation of features.

Return type

TrainValidIterator

Returns

Copy of object with transformed features.

apply_selector(selector)[source]

Select features on train data.

Check if selector is fitted. If not - fit and then perform selection. If fitted, check if it’s ok to apply.

Parameters

selector – Uses for feature selection.

Return type

TrainValidIterator

Returns

Dataset with selected features.

convert_to_holdout_iterator()[source]

Abstract method. Convert iterator to HoldoutIterator.

Return type

HoldoutIterator