"min max ex [log (d (x))]+ez [log (1-d (g (z))] is used to show the actual algorithm for creating it.
The portrait of Edmund Bellamy was created by an art group in Paris using the AI method of deep learning. This kind of deep learning is a generative antagonistic network (GAN) model, which produces results through the interaction between generative model and discriminant model. The creative team first shot15,000 classic portraits from 14 to the 20th century to "train" the computer to "understand" the "creative law" of the portraits, establish algorithm formulas, and form a "generation model" to generate images. Then use the computer's "discriminant model" to find out the difference between the computer-generated image and the real painting and make a judgment. The algorithm of the generated model is constantly modified until it is difficult for the generated image to judge whether it is a real painting or not, that is, the computer will "learn something" and use the algorithm formula of the generated model to "create".