8d4fee5d7dc69f82aadb38abc4564dda5cc8580f,rankeval/core/metrics/rbp.py,RBP,eval_per_query,#RBP#Any#Any#,58

Before Change

        // expected_utility
        expected_utility = 0.
        binary_rel = y[idx_y_pred_sorted] > self.threshold
        for i, idx in enumerate(idx_y_pred_sorted):
            expected_utility += binary_rel[i] * pow(self.p, i)

        rbp = (1. - self.p) * expected_utility
        return rbp

After Change

        if self.cutoff is not None:
            idx_y_pred_sorted = idx_y_pred_sorted[:self.cutoff]

        discount = np.power(self.p, np.arange(len(idx_y_pred_sorted)))
        gain = y[idx_y_pred_sorted] > self.threshold

        rbp = (1. - self.p) * (gain * discount).sum()
        return rbp

Instances