def __init__()

in ludwig/utils/metrics_utils.py [0:0]

• 23 lines of code
• 12 McCabe index (conditional complexity)

    def __init__(self, conditions, predictions, labels=None,
                 sample_weight=None):
        # assert (len(predictions) == len(conditions))
        min_length = min(len(predictions), len(conditions))
        self.predictions = predictions[:min_length]
        self.conditions = conditions[:min_length]

        if labels is not None:
            self.label2idx = {label: idx for idx, label in enumerate(labels)}
            self.idx2label = {idx: label for idx, label in enumerate(labels)}
            labels = list(range(len(labels)))
        else:
            self.label2idx = {str(label): idx for idx, label in
                              enumerate(np.unique(
                                  [self.predictions, self.conditions]))}
            self.idx2label = {idx: str(label) for idx, label in
                              enumerate(np.unique(
                                  [self.predictions, self.conditions]))}
        self.cm = confusion_matrix(self.conditions,
                                   self.predictions,
                                   labels=labels,
                                   sample_weight=sample_weight)

        # if labels is not None:
        #     self.labels_dict = {label: idx for idx, label in enumerate(labels)}
        # else:
        #     if conditions.dtype.char == 'S':  # it's an array of strings
        #         self.labels_dict = {str(label): idx for idx, label in
        #                             enumerate(np.unique([predictions, conditions]))}
        #     else:  # numerical
        #         max_label = np.concatenate([predictions, conditions]).max()
        #         self.labels_dict = {str(i): i for i in range(max_label + 1)}
        #         labels = [str(i) for i in range(max_label + 1)]
        # self.cm = confusion_matrix(conditions, predictions, labels, sample_weight)

        self.sum_predictions = np.sum(self.cm, axis=0)
        self.sum_conditions = np.sum(self.cm, axis=1)
        self.all = np.sum(self.cm)