DCGAN is initialized with random weights, so a random code plugged into the network would deliver a very random impression. Even so, while you might imagine, the network has millions of parameters that we can easily tweak, plus the purpose is to find a location of those parameters that makes samples generated from random codes appear to be the scho