def score(self, X: np.ndarray, Y: np.ndarray, k: int = 2, dist_type: str = 'point') \
            -> Tuple[np.ndarray, np.ndarray]:

  d = np.tile(None, (X.shape[0], self.classes))  # init distance matrix: [nb instances, nb classes]

  for c in range(self.classes):
      d_tmp = self.kdtrees[c].query(X, k=k)[0]  # get k nearest neighbors for each class
      if dist_type == 'point':
          d[:, c] = d_tmp[:, -1]
      elif dist_type == 'mean':
          d[:, c] = np.mean(d_tmp, axis=1)

  sorted_d = np.sort(d, axis=1)  # sort distance each instance in batch over classes
  # get distance to predicted and closest other class and calculate trust score
  d_to_pred = d[range(d.shape[0]), Y]
  d_to_closest_not_pred = np.where(sorted_d[:, 0] != d_to_pred, sorted_d[:, 0], sorted_d[:, 1])
  trust_score = d_to_closest_not_pred / (d_to_pred + self.eps)
  # closest not predicted class
  class_closest_not_pred = np.where(d == d_to_closest_not_pred.reshape(-1, 1))[1]
  return trust_score, class_closest_not_pred