DCGAN is initialized with random weights, so a random code plugged in the network would make a totally random picture. Even so, while you might imagine, the network has countless parameters that we can tweak, and also the goal is to locate a placing of those parameters that makes samples generated from random codes appear to be the teaching knowled