DCGAN is initialized with random weights, so a random code plugged in the network would generate a completely random impression. However, while you may think, the network has many parameters that we are able to tweak, and the target is to locate a setting of these parameters that makes samples produced from random codes appear like the schooling da