DCGAN is initialized with random weights, so a random code plugged into your network would make a completely random image. Having said that, when you may think, the network has millions of parameters that we can tweak, and also the aim is to locate a location of such parameters which makes samples produced from random codes look like the coaching d