Total loss

We just learned the loss function of the generator and the discriminator combining these two losses, and we write our final loss function as follows:

So, our objective function is basically a min-max objective function, that is, a maximization for the discriminator and a minimization for the generator, and we find the optimal generator parameter, , and discriminator parameter, , through backpropagating the respective networks.

So, we perform gradient ascent; that is, maximization on the discriminator:

And, we perform gradient descent; that is, minimization on the generator: