Understanding of Metric for Precision and Recall for Multi-label Classification

[pengpaiSH]

Keras provides a metric of precision and recall for multi-label classification. The document says that "Calculates the precision, a metric for multi-label classification of how many selected items are relevant." And the source code in /keras/metrics.py:

def precision(y_true, y_pred):
    """Precision metric.
    Only computes a batch-wise average of precision.
    Computes the precision, a metric for multi-label classification of
    how many selected items are relevant.
    """
    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
    predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
    precision = true_positives / (predicted_positives + K.epsilon())
    return precision

If I understand correctly, precision will count each element as a single label. In other words, if y_true and y_pred are 2D arrays, e.g. with shape (32, 13) if the batch size is 32 and the number of classes is 13. What if I only care about class-level precision? For example, if I would like to know the 1st class precision, 2nd class precision ... etc ?

[callicles]

Hi,

Those metrics have been removed in 2.0. #5794

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 30 days if no further activity occurs, but feel free to re-open a closed issue if needed.