August 2019
Intermediate to advanced
342 pages
9h 35m
English
We have said that the purpose of a GAN is to achieve a condition of equilibrium between the two NNs. The search for this equilibrium involves solving the following equation, a minimax condition:

From the preceding formula, you can see the antagonistic goal that characterizes the two neural networks. We try to maximize D while minimizing G. In other words, the neural network D, which represents the discriminator, aims to maximize the equation, which translates into maximizing the output associated with real samples while minimizing the output associated with fake samples. On the other side, the neural network G, which represents ...
Read now
Unlock full access