Unused hyperparameters remain active when datasets are purely categorical or purely numerical

[gui-miotto]

PR #735 implements data preprocessing of categorical and numerical features as two parallel pipelines, each with its own independent hyperparameters.
Ideally, if the dataset is made just of features of the same kind, then just the hyperparameters of the corresponding pipeline should be considered for optimization.
Nevertheless, this is not the current behavior. At the moment, the hyperparameters of both pipelines remain active regardless of the dataset.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs for the next 7 days. Thank you for your contributions.

[Louquinze]

Can we exploit the include or exclude option of autosklearn for fixing this ?
This parameters are providing the possibility to enable or disable certain steps of the preprocessing.

As i understand this also means not including this steps in the search space.
see auto-sklearn/autosklearn/automl.py line 2095 ff.

and auto-sklearn/autosklearn/util/pipeline.py:get_configuration_space

[mfeurer]

Can we exploit the include or exclude option of autosklearn for fixing this ?

Unfortunately not. For this the user would need to know what to include or exclude based on the given data type, which is something we cannot expect.