Personalized Federated Learning with Semisupervised Distillation

Algorithm 2

Server execute.

 Input: number of clients , number of communication rounds

 Output: Aggregated prediction

 Assign public unlabeled datasets

 for each round do

   select random clients from

  ifthen

   fordo

     ClientUpdate(t)

    Compute by Equation (4)

   end

  else

   fordo

     ClientUpdate( t)

    Compute by Equation (4)

   end

 end

  Compute by equation (5)

 end