(Single choice question) The following description of generative adversarial networks is incorrect () . A、Generative adversarial network consists of two parts: generator and discriminator B、When the discriminator of adversarial network is training, its input is the image generated by generator and the real image from the training set. C、GAN's Generators generate images from random noise (usually taken from a uniform or Gaussian distribution) D、Since generative adversarial networks are unsupervised models, no training data is required