Hands-On Generative Adversarial Networks with Keras

Least Squares GAN

The Least Squares GAN uses the least squares objective function to train the discriminator and generator. Unlike the Sigmoid Cross Entropy, the Least Squares loss more heavily penalizes samples regarding their position with respect to the decision boundary. In their paper, the authors affirm that the LSGAN contributes to the stability of the learning process, removes the need of using batch normalization, and converges faster than the Wasserstein GAN.

The objective functions of LSGAN for the discriminator D and generator G are the following:

The authors of LSGAN state that, with specific relationships between a, b, and c ...