DCGAN is initialized with random weights, so a random code plugged into the network would produce a totally random image. Nonetheless, while you may think, the network has a lot of parameters that we can easily tweak, along with the target is to locate a location of such parameters that makes samples generated from random codes look like the coachi