some question about the sfl

[a1ei]

Hello，the scenario described in your article is SL, but the code is federated learning, which involves parameter aggregation. Is there any difference in this? Or can it be considered as SL when the client is 1? If so, is the update of the trunk also local

[StupidTrees]

Yes, the situation in the article is implemented with a client number of 1.

The split learning framework implemented in this repository is actually Federated Split Learning (FSL). That is, when the number of clients exceeds 1, the parameter aggregation method is employed instead of training the clients sequentially. In the implementation, the server needs to maintain an update for the trunk part of each client.