Different neural network architectures at the parties

[CriZ-sd]

Hi,
does IBM FL support to use different Neural Network architecture at each party(for example using 4 parties and each party trains a different local Neural Network architecture)? If yes, then which fusion algorithm (aggregation protocol) is better for this purpose ?

[Yi-Zoey]

Hi @CriZ-sd you can check and see if PFNM works for your use case? Also, see the paper for PFNM here for more information.