File size: 11,255 Bytes
d2a8669
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
import numpy as np
from warnings import warn

from aif360.algorithms import Transformer
from aif360.metrics import utils
from aif360.metrics import BinaryLabelDatasetMetric, ClassificationMetric


class RejectOptionClassification(Transformer):

    """Reject option classification is a postprocessing technique that gives
    favorable outcomes to unpriviliged groups and unfavorable outcomes to
    priviliged groups in a confidence band around the decision boundary with the
    highest uncertainty [10]_.

    References:
        .. [10] F. Kamiran, A. Karim, and X. Zhang, "Decision Theory for
           Discrimination-Aware Classification," IEEE International Conference
           on Data Mining, 2012.
    """

    def __init__(self, unprivileged_groups, privileged_groups,
                low_class_thresh=0.01, high_class_thresh=0.99,
                num_class_thresh=100, num_ROC_margin=50,
                metric_name="Statistical parity difference",
                metric_ub=0.05, metric_lb=-0.05):
        """
        Args:
            unprivileged_groups (dict or list(dict)): Representation for
                unprivileged group.
            privileged_groups (dict or list(dict)): Representation for
                privileged group.
            low_class_thresh (float): Smallest classification threshold to use
                in the optimization. Should be between 0. and 1.
            high_class_thresh (float): Highest classification threshold to use
                in the optimization. Should be between 0. and 1.
            num_class_thresh (int): Number of classification thresholds between
                low_class_thresh and high_class_thresh for the optimization
                search. Should be > 0.
            num_ROC_margin (int): Number of relevant ROC margins to be used in
                the optimization search. Should be > 0.
            metric_name (str): Name of the metric to use for the optimization.
                Allowed options are "Statistical parity difference",
                "Average odds difference", "Equal opportunity difference".
            metric_ub (float): Upper bound of constraint on the metric value
            metric_lb (float): Lower bound of constraint on the metric value
        """
        super(RejectOptionClassification, self).__init__(
            unprivileged_groups=unprivileged_groups,
            privileged_groups=privileged_groups,
            low_class_thresh=low_class_thresh, high_class_thresh=high_class_thresh,
            num_class_thresh=num_class_thresh, num_ROC_margin=num_ROC_margin,
            metric_name=metric_name)

        allowed_metrics = ["Statistical parity difference",
                           "Average odds difference",
                           "Equal opportunity difference"]

        self.unprivileged_groups = unprivileged_groups
        self.privileged_groups = privileged_groups

        self.low_class_thresh = low_class_thresh
        self.high_class_thresh = high_class_thresh
        self.num_class_thresh = num_class_thresh
        self.num_ROC_margin = num_ROC_margin
        self.metric_name = metric_name
        self.metric_ub = metric_ub
        self.metric_lb = metric_lb

        self.classification_threshold = None
        self.ROC_margin = None

        if ((self.low_class_thresh < 0.0) or (self.low_class_thresh > 1.0) or\
            (self.high_class_thresh < 0.0) or (self.high_class_thresh > 1.0) or\
            (self.low_class_thresh >= self.high_class_thresh) or\
            (self.num_class_thresh < 1) or (self.num_ROC_margin < 1)):

            raise ValueError("Input parameter values out of bounds")

        if metric_name not in allowed_metrics:
            raise ValueError("metric name not in the list of allowed metrics")

    def fit(self, dataset_true, dataset_pred):
        """Estimates the optimal classification threshold and margin for reject
        option classification that optimizes the metric provided.

        Note:
            The `fit` function is a no-op for this algorithm.

        Args:
            dataset_true (BinaryLabelDataset): Dataset containing the true
                `labels`.
            dataset_pred (BinaryLabelDataset): Dataset containing the predicted
                `scores`.

        Returns:
            RejectOptionClassification: Returns self.
        """

        fair_metric_arr = np.zeros(self.num_class_thresh*self.num_ROC_margin)
        balanced_acc_arr = np.zeros_like(fair_metric_arr)
        ROC_margin_arr = np.zeros_like(fair_metric_arr)
        class_thresh_arr = np.zeros_like(fair_metric_arr)

        cnt = 0
        # Iterate through class thresholds
        for class_thresh in np.linspace(self.low_class_thresh,
                                        self.high_class_thresh,
                                        self.num_class_thresh):

            self.classification_threshold = class_thresh
            if class_thresh <= 0.5:
                low_ROC_margin = 0.0
                high_ROC_margin = class_thresh
            else:
                low_ROC_margin = 0.0
                high_ROC_margin = (1.0-class_thresh)

            # Iterate through ROC margins
            for ROC_margin in np.linspace(
                                low_ROC_margin,
                                high_ROC_margin,
                                self.num_ROC_margin):
                self.ROC_margin = ROC_margin

                # Predict using the current threshold and margin
                dataset_transf_pred = self.predict(dataset_pred)

                dataset_transf_metric_pred = BinaryLabelDatasetMetric(
                                             dataset_transf_pred,
                                             unprivileged_groups=self.unprivileged_groups,
                                             privileged_groups=self.privileged_groups)
                classified_transf_metric = ClassificationMetric(
                                             dataset_true,
                                             dataset_transf_pred,
                                             unprivileged_groups=self.unprivileged_groups,
                                             privileged_groups=self.privileged_groups)

                ROC_margin_arr[cnt] = self.ROC_margin
                class_thresh_arr[cnt] = self.classification_threshold

                # Balanced accuracy and fairness metric computations
                balanced_acc_arr[cnt] = 0.5*(classified_transf_metric.true_positive_rate()\
                                       +classified_transf_metric.true_negative_rate())
                if self.metric_name == "Statistical parity difference":
                    fair_metric_arr[cnt] = dataset_transf_metric_pred.mean_difference()
                elif self.metric_name == "Average odds difference":
                    fair_metric_arr[cnt] = classified_transf_metric.average_odds_difference()
                elif self.metric_name == "Equal opportunity difference":
                    fair_metric_arr[cnt] = classified_transf_metric.equal_opportunity_difference()

                cnt += 1

        rel_inds = np.logical_and(fair_metric_arr >= self.metric_lb,
                                  fair_metric_arr <= self.metric_ub)
        if any(rel_inds):
            best_ind = np.where(balanced_acc_arr[rel_inds]
                                == np.max(balanced_acc_arr[rel_inds]))[0][0]
        else:
            warn("Unable to satisy fairness constraints")
            rel_inds = np.ones(len(fair_metric_arr), dtype=bool)
            best_ind = np.where(fair_metric_arr[rel_inds]
                                == np.min(fair_metric_arr[rel_inds]))[0][0]

        self.ROC_margin = ROC_margin_arr[rel_inds][best_ind]
        self.classification_threshold = class_thresh_arr[rel_inds][best_ind]

        return self

    def predict(self, dataset):
        """Obtain fair predictions using the ROC method.

        Args:
            dataset (BinaryLabelDataset): Dataset containing scores that will
                be used to compute predicted labels.

        Returns:
            dataset_pred (BinaryLabelDataset): Output dataset with potentially
            fair predictions obtain using the ROC method.
        """
        dataset_new = dataset.copy(deepcopy=False)

        fav_pred_inds = (dataset.scores > self.classification_threshold)
        unfav_pred_inds = ~fav_pred_inds

        y_pred = np.zeros(dataset.scores.shape)
        y_pred[fav_pred_inds] = dataset.favorable_label
        y_pred[unfav_pred_inds] = dataset.unfavorable_label

        # Indices of critical region around the classification boundary
        crit_region_inds = np.logical_and(
                dataset.scores <= self.classification_threshold+self.ROC_margin,
                dataset.scores > self.classification_threshold-self.ROC_margin)

        # Indices of privileged and unprivileged groups
        cond_priv = utils.compute_boolean_conditioning_vector(
                        dataset.protected_attributes,
                        dataset.protected_attribute_names,
                        self.privileged_groups)
        cond_unpriv = utils.compute_boolean_conditioning_vector(
                        dataset.protected_attributes,
                        dataset.protected_attribute_names,
                        self.unprivileged_groups)

        # New, fairer labels
        dataset_new.labels = y_pred
        dataset_new.labels[np.logical_and(crit_region_inds,
                            cond_priv.reshape(-1,1))] = dataset.unfavorable_label
        dataset_new.labels[np.logical_and(crit_region_inds,
                            cond_unpriv.reshape(-1,1))] = dataset.favorable_label

        return dataset_new

    def fit_predict(self, dataset_true, dataset_pred):
        """fit and predict methods sequentially."""
        return self.fit(dataset_true, dataset_pred).predict(dataset_pred)

# Function to obtain the pareto frontier
def _get_pareto_frontier(scores, return_mask = True):  # <- Fastest for many points
    """
    :param scores: An (n_points, n_scores) array
    :param return_mask: True to return a mask, False to return integer indices of efficient points.
    :return: An array of indices of pareto-efficient points.
        If return_mask is True, this will be an (n_points, ) boolean array
        Otherwise it will be a (n_efficient_points, ) integer array of indices.

    adapted from: https://stackoverflow.com/questions/32791911/fast-calculation-of-pareto-front-in-python
    """
    is_efficient = np.arange(scores.shape[0])
    n_points = scores.shape[0]
    next_point_index = 0  # Next index in the is_efficient array to search for

    while next_point_index<len(scores):
        nondominated_point_mask = np.any(scores>=scores[next_point_index], axis=1)
        is_efficient = is_efficient[nondominated_point_mask]  # Remove dominated points
        scores = scores[nondominated_point_mask]
        next_point_index = np.sum(nondominated_point_mask[:next_point_index])+1

    if return_mask:
        is_efficient_mask = np.zeros(n_points, dtype = bool)
        is_efficient_mask[is_efficient] = True
        return is_efficient_mask
    else:
        return is_efficient