Energy-based Generative Adversarial Network

We introduce the "Energy-based Generative Adversarial Network'" model (EBGAN) which views the discriminator as an energy function that associates low energies with the regions near the data manifold and higher energies with other regions. Similar to the probabilistic GANs, a generator is trained to produce contrastive samples with minimal energies, while the discriminator is trained to assign high energies to these generated samples. Viewing the discriminator as an energy function allows to use a wide variety of architectures and loss functionals in addition to the usual binary classifier with logistic output. Among them, an instantiation of EBGAN is to use an auto-encoder architecture, with the energy being the reconstruction error. We show that this form of EBGAN exhibits more stable behavior than regular GANs during training. We also show that a single-scale architecture can be trained to generate high-resolution images.