armlet.data.cleaning package

Subpackages

• armlet.data.cleaning.missing_values package
  • Module contents
    • MissingValuesDataCleaningMethod
      • MissingValuesDataCleaningMethod.clean_errors_dataloader()
      • MissingValuesDataCleaningMethod.count_n_detected_errors_by_sensitive_attributes()
      • MissingValuesDataCleaningMethod.handle_missing_values()
    • remove_missing_values()

Module contents

This module contains the components for performing data cleaning.

class armlet.data.cleaning.DataCleaningMethod(sensitive_attributes: list[str], error_type)

Bases: ABC

clean_client_errors(clients_data_dict)

clean_errors(all_data)

abstractmethod clean_errors_dataloader(data: tuple[DataFrame, DataFrame]) → tuple[DataFrame, DataFrame]

clean_server_errors(server_data)

compute_cleaning_metrics()

count_n_samples_by_sensitive_attributes(X)

init_counters()

armlet.data.cleaning.compute_global_stat_values(clients_data_dict: dict[str, tuple[DataFrame, DataFrame]], cfg)

armlet.data.cleaning.data_cleaning_pipeline(data, cfg_cleaning, sensitive_attributes)

Clean the data by following the data cleaning pipeline.

Warning

Label errors and missing values cleaning methods are only applicable to tabular data. Missing values cleaning methods can also be used with image datasets.