PR Summary

This PR enhances the MutualInformation function by adding additional validations to ensure the dataset contains the necessary numeric features and a target column before proceeding with mutual information calculations. In detail, the following changes have been made:

• Added a check that raises an exception if the dataset does not have any numeric features (dataset.feature_columns_numeric), ensuring that the calculation only proceeds when valid numeric features are available.
• Introduced a check to ensure that a target column is provided in the dataset (dataset.target_column) before performing computation.
• Modified how the feature and target variables (X and y) are retrieved from the dataset, now obtaining values from the internal dataframe (dataset._df) using the specified column identifiers.
• The selection of the mutual information function (classification vs. regression) remains intact, assuming that the task input is valid.

These changes aim to improve the robustness of the function by providing clear error messages when essential dataset properties are missing.

Test Suggestions

• Write tests where dataset.feature_columns_numeric is empty, and verify that the correct ValueError is raised with the expected message.
• Write tests where dataset.target_column is not set, and verify that the function raises the appropriate ValueError.
• Test the function with a valid dataset that includes numeric features and a target column to ensure that the mutual information calculation proceeds without error.
• Test both 'classification' and 'regression' task pathways to ensure correct behavior for valid inputs.