4abda80dd12518e6bfdc44d067566a3e6947e906,deslib/dcs/a_priori.py,APriori,estimate_competence,#APriori#Any#Any#,91

Before Change

        for clf_index in range(self.n_classifiers):

            // Check if the dynamic frienemy pruning (DFP) should be used used
            if self.DFP_mask[clf_index]:
                result = np.zeros(self.k)
                for counter, index in enumerate(idx_neighbors):
                    target = self.DSEL_target[index]
                    // get the post_prob for the correct class
                    //post_prob = self._get_scores_dsel(clf_index, index)[target]
                    post_prob = self.dsel_scores[index, clf_index, target]

                    result[counter] = (post_prob * dists_normalized[counter])

                competences[clf_index] = sum(result)/sum(dists_normalized)
        return competences

After Change

                      The competence level estimated for each base classifier
        
        dists, idx_neighbors = self._get_region_competence(query)
        idx_neighbors = np.atleast_2d(idx_neighbors)
        dists = np.atleast_2d(dists)
        dists_normalized = 1.0/dists

        // Get the ndarray containing the scores obtained for the correct class for each neighbor (and test sample)
        scores_target_class = self.dsel_scores[idx_neighbors, :, self.DSEL_target[idx_neighbors]]

        // Multiply the scores obtained for the correct class to the distances of each corresponding neighbor
        scores_target_class *= np.expand_dims(dists_normalized, axis=2)

        // Sum the scores obtained for each neighbor and divide by the sum of all distances
        competences = np.sum(scores_target_class, axis=1)/ np.sum(dists_normalized, axis=1)

Instances