holisticai.robustness.metrics.adversarial_accuracy

holisticai.robustness.metrics.adversarial_accuracy(y, y_pred, y_adv_pred)

Calculate the adversarial accuracy of a model with array-like inputs.

Parameters

yarray-like

The true labels. If None, the function calculates the accuracy based on y_pred and y_adv_pred directly.

y_predarray-like

The predicted labels for the original input.

y_adv_predarray-like

The predicted labels for the adversarial input.

Returns

float

The adversarial accuracy value.

Examples

>>> import numpy as np
>>> from holisticai.robustness.metrics import adversarial_accuracy
>>> y = [1, 0, 1, 1]  # Example with list input
>>> y_pred = [1, 0, 0, 1]
>>> y_adv_pred = [0, 0, 1, 1]
>>> adversarial_accuracy(y, y_pred, y_adv_pred)
0.6666666666666666