Comment by gat1

I guess we do not know anything about the training dataset ?