Hi, everyone, I am a sutdent and working on federated learning communication efficiency. Currently, I am preparing to implement a three layer FL training using flower, as shown in the picture.
For the edge layer to middle layer, I want to use a aggregation method
AggreAlog 1, and for the middle layer to the clould layer, I want to use another aggregation method AggreAlgo 2.I have searched from the slack channels and found that several questions about it, but I didn’t found a feasible solution. The replies I see said flower is capable of implementing this scenrio, but I didn’t found any code example for referencing. So, is there any related codes for referencing?
When I walking through the documentaton of flower, I found
clieanmanager and clientproxy may be helpful to implemnet the scenario, but I don’t know if it can work. So, is clientmanager and clientproxy can be a solution for the implementation of this scenario？