DCGAN is initialized with random weights, so a random code plugged in the network would crank out a totally random impression. Having said that, when you may think, the network has many parameters that we can tweak, and the goal is to locate a location of such parameters which makes samples produced from random codes seem like the coaching informat