holisticai.bias.mitigation.PluginEstimationAndCalibration#

class holisticai.bias.mitigation.PluginEstimationAndCalibration(*args, **kargs)[source]#

Plugin Estimation and Calibration postprocessing [1] optimizes over calibrated regressor outputs via a smooth optimization. The rates of convergence of the proposed estimator were derived in terms of the risk and fairness constraint.

Parameters

Lint: The number of thresholds to estimate.
betafloat: The regularization parameter

Examples

>>> from holisticai.bias.mitigation import PluginEstimationAndCalibration
>>> mitigator = PluginEstimationAndCalibration()
>>> mitigator.fit_transform(y_pred, group_a, group_b)
>>> test_data_transformed = mitigator.transform(y_pred, group_a, group_b)

References

fit(y_pred: ndarray, group_a: ndarray, group_b: ndarray)[source]#

Compute a fair regression function by minimizing the squared error subject to a fairness constraint.

Description

Compute a fair predictor by estimating a regression function by standard methods and then estimate the thresholds to solve the minimization problem.

Parameters

y_predarray-like: Predicted vector (num_examples, ).
group_aarray-like: Group membership vector (binary)
group_barray-like: Group membership vector (binary)

Returns

Self

fit_transform(y_pred: ndarray, group_a: ndarray, group_b: ndarray)[source]#

Fit and transform

Description

Fit and transform

Parameters

y_predarray-like: Predicted vector (num_examples,).
group_aarray-like: Group membership vector (binary)
group_barray-like: Group membership vector (binary)

Returns

dict: A dictionary with new predictions

transform(y_pred: ndarray, group_a: ndarray, group_b: ndarray)[source]#

Apply transform function to predictions and likelihoods

Description

Use a fitted probability to change the output label and invert the likelihood

Parameters

y_predarray-like: Predicted vector (nb_examples,)
group_aarray-like: Group membership vector (binary)
group_barray-like: Group membership vector (binary)
thresholdfloat: float value to discriminate between 0 and 1

Returns

dict: A dictionary with new predictions