Whether they are or not is irrelevant to the difference- there is still an explicit division between training and inference, with reinforcement of the statistic

Whether they are or not is irrelevant to the difference- there is still an explicit division between training and inference, with reinforcement of the statistics of the training set.