Question on federated unsupervised domain adaptation

[ccc202020]

Firstly, thank for the excellent work you have contributed to the privacy protection clinical set.
But, I have some question to ask so that I can know more clear about the federated learning.
I want to know that in federated unsupervised domain adaptation setting, we have two different classifier(local and global), the local classifier is trained using one site data and the global classifier is updated based on the four different local classifiers. For example, the NYU site data have been splited into train and test set, the global classifier will contain some information of NYU site(for the using NYU train set to train the local classifier and the global classifier is updated based on local classifier). If it is a semi-supervised learning and not a unsupervised learning setting.
I want to know if I have understand the federated learning correctly, so I wish to get your reply sincerely
Finally, thanks for your excellent work on the federated domain adaptation one more time!