DCGAN is initialized with random weights, so a random code plugged to the network would deliver a totally random impression. Even so, while you might imagine, the network has countless parameters that we can easily tweak, plus the target is to find a setting of these parameters which makes samples