Loss computation in pfedgraph_cosine.py

[tesiawang]

Hi there, thanks a lot for the open-source codes!
Could you please tell me why 'args.lam' is positive in pfedgraph_cosine.py? This hyper-parameter is multiplied with loss2, which actually is the similarity instead of distance. Thus I personally think args.lam should be negative, just as what the equation (1) in the paper indicates.