DCGAN is initialized with random weights, so a random code plugged in the network would crank out a totally random graphic. However, when you may think, the network has millions of parameters that we could tweak, plus the purpose is to find a setting of these parameters which makes samples created from random codes appear like the instruction infor