Artificial Intelligence-Based Digital Media Design Effect Enhancement Mechanism

Algorithm 1

The training algorithm of GAN.

	Input: raw data and noise data
	Output: The GAN model
(1)	initialize the parameters and of the model randomly;
(2)	for each epoch i = 1, 2, …, do
(3)	for each step j = 1, 2, …, do
(4)	sample minibatch of m noise samples from noise data;
(5)	sample minibatch of m examples x from raw data;
(6)	calculate the loss and gradient:
(7)	update the parameters and of the discriminator D;
(8)	end for
(9)	sample minibatch of m noise samples from noise data;
(10)	calculate the partial derivatives of the parameters:
(11)	update the parameters and of the generator G;
(12)	end for
(13)	Return the GAN model with parameters.