Least Squares Generative Adversarial Networks

ABSTRACT

Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. This loss function, however, may lead to the vanishing gradient problem during the learning process. To overcome such problem, here we propose the Least Squares Generative Adversarial Networks (LSGANs) that adopt the least squares loss function for the discriminator. We show that minimizing the objective function of LSGAN yields minimizing the Pearson $χ^2$ divergence. There are two benefits of LSGANs over regular GANs. First, LSGANs are able to generate higher quality images than regular GANs. Second, LSGANs performs more stable during the learning process. We evaluate the LSGANs on five scene datasets and the experimental results demonstrate that the generated images by LSGANs look more realistic than the ones generated by regular GANs. We also conduct two comparison experiments between LSGANs and regular GANs to illustrate the stability of LSGANs.