The GAN is dead; long live the GAN! A Modern GAN Baseline

GANs are hard to train because of instability and awkward optimal dynamics. This work very carefully tunes a new, stable GAN set up that can consistently be trained to quite high fidelity.

#technology #ai #gan