DCGAN is initialized with random weights, so a random code plugged to the network would deliver a completely random picture. However, as you may think, the network has countless parameters that we will tweak, and the objective is to find a setting of those parameters that makes samples created from random codes appear like the training facts.Prompt