In the dataset description, we saw that the training data is "automatically generated", hence "noisy". Does that mean the labels in the training set are completely random? Or is it generated according to some criteria?
Posted by: Tetheras @ March 21, 2018, 2:33 p.m.