1: $.  In comparison with the (centralized) gradient method of the convergence rate $