gaussianconditionalsufficientstats.py 文件源码

def _get_split_point_candidates(self):
        splits = SortedList()
        min_value = math.inf
        max_value = -math.inf

        for class_val, att_estimator in self._class_lookup.items():
            min_val_observed_for_class_val = self._min_val_observed_per_class.get(class_val, None)
            if min_val_observed_for_class_val is not None:
                if min_val_observed_for_class_val < min_value:
                    min_value = min_val_observed_for_class_val
                max_val_observed_for_class_val = self._max_val_observed_per_class.get(class_val)
                if max_val_observed_for_class_val > max_value:
                    max_value = max_val_observed_for_class_val

        if min_value < math.inf:
            new_bin = max_value - min_value
            new_bin /= (self._num_bins + 1)
            for i in range(self._num_bins):
                split = min_value + (new_bin * (i + 1))
                if split > min_value and split < max_value:
                    splits.add(split)
        return splits