LFR

Learning fair representations is a pre-processing technique that finds a latent representation which encodes the data well but obfuscates information about protected attributes [2]. .. rubric:: References

[2]

R. Zemel, Y. Wu, K. Swersky, T. Pitassi, and C. Dwork, “Learning Fair Representations.” International Conference on Machine Learning, 2013.

Based on code from https://github.com/zjelveh/learning-fair-representations

class fairdiverse.search.preprocessing_model.LFR.LFR(configs, dataset)

Bases: PreprocessingFairnessIntervention

Learning Fair Representations (LFR) fairness intervention.

This class applies the LFR approach to modify the dataset such that fairness constraints are met while preserving as much utility as possible.

fit(X_train, run)

Train the LFR fairness model using the given training dataset.

This method optimizes a fairness objective by learning fair representations of the data using constrained optimization.

:param X_trainpandas.DataFrame or numpy.ndarray

The training dataset. It is assumed that the last non-sensitive column is the target variable.

:param runstr

The identifier for the training run.

:returnself

The trained LFR model.

transform(X, run, file_name=None)

Apply the fairness transformation to the dataset using the learned model.

This method ensures fairness by adjusting feature distributions while maintaining data utility.

:param Xpandas.DataFrame

The dataset to which the fairness transformation is applied.

:param runstr

The identifier for the transformation run.

:param file_namestr, optional

Name of the file to save the transformed dataset.

:returnpandas.DataFrame

The dataset with transformed fair columns.