In the paper,you viewed the real and fake images equally,as to the loss function, when defining loss function , why you considered it into two parts ?