gemseo / mlearning / resampling

Hide inherited members

base_resampler module

A base class for resampling and surrogate modeling.

class gemseo.mlearning.resampling.base_resampler.BaseResampler(sample_indices, n_splits, seed=0)[source]

Bases: object

A base class for resampling and surrogate modeling.

Parameters:

  • sample_indices (NDArray[int]) – The original indices of the samples.

  • n_splits (int) – The number of train-test splits.

  • seed (int | None) –

    The seed to initialize the random generator. If None, then fresh, unpredictable entropy will be pulled from the OS.

    By default it is set to 0.

execute(model, return_models=False, input_data=None, stack_predictions=True, fit_transformers=True, store_sampling_result=False)[source]

Apply the resampling technique to a machine learning model.

Parameters:

  • model (BaseMLAlgo) – The machine learning model.

  • return_models (bool) –

    Whether the sub-models resulting from resampling are returned.

    By default it is set to False.

  • input_data (ndarray | None) – The input data for the prediction, if any.

  • stack_predictions (bool) –

    Whether the sub-predictions are stacked per sub-model (first the predictions of the first sub-model, then the prediction of the second sub-model, etc.). This argument is ignored when input_data is None.

    By default it is set to True.

  • fit_transformers (bool) –

    Whether to re-fit the transformers.

    By default it is set to True.

  • store_sampling_result (bool) –

    Whether to store the sampling results in the attribute resampling_results of the original model.

    By default it is set to False.

Returns:

First the sub-models resulting from resampling if return_models is True then the predictions, either per fold or stacked.

Raises:

ValueError – When the model is neither a supervised algorithm nor a clustering one.

Return type:

tuple[list[BaseMLAlgo], list[ndarray] | ndarray]

plot(file_path='', show=True, colors=('b', 'r'))[source]

Plot the train-test splits.

Parameters:

  • file_path (str | Path) –

    The file path to save the figure. If empty, do not save the figure.

    By default it is set to “”.

  • show (bool) –

    Whether to display the figure.

    By default it is set to True.

  • colors (tuple[str, str]) –

    The colors for training and test points.

    By default it is set to (‘b’, ‘r’).

Returns:

The visualization.

Return type:

Scatter

name: str

The name of the resampler.

Use the class name by default.

property sample_indices: NDArray[int]

The indices of the samples after shuffling.

property seed: int

The seed to initialize the random generator.

property splits: Splits

The train-test splits resulting from the splitting of the samples.

A train-test split is a partition whose first component contains the indices of the learning samples and the second one the indices of the test samples.