DCGAN is initialized with random weights, so a random code plugged in the network would produce a totally random picture. However, as you may think, the network has countless parameters that we can tweak, and also the intention is to find a placing of such parameters that makes samples generated from random codes seem like the training knowledge.In