DCGAN is initialized with random weights, so a random code plugged into the network would deliver a completely random picture. Nonetheless, as you may think, the network has many parameters that we could tweak, along with the aim is to find a location of those parameters that makes samples created from random codes appear like the coaching data.We