Labels for unlabeled data about mc-net HOT 3 CLOSED

That would be fine

from mc-net.

Random labels cannot evaluate the performance of given unlabeled data. You can only output the predictions and ignore the performance evaluation part. If you apply random labels, the obtained performance is not reliable. Thanks.

from mc-net.

Thank you, what if I divide my labeled data 80/20 and train my model with 80% labeled + unlabeled data and then evaluate on remaining 20?

from mc-net.