arXiv (Cornell University) · 2016 · 895 citations · 19 references
ComputingComputer ScienceComputer VisionFeature EngineeringMachine LearningDeep LearningGenerative ArchitectureGenerative ModelingImage SynthesisNeural NetworksAdversarial LossAutoencodersEnergy-based Generative ModelsEnergy-based ModelsGenerative Adversarial NetworksGenerative Image SynthesisLogistic LossManifold Learning
Similar to probabilistic GANs, the generator produces contrastive samples with minimal energies while the discriminator assigns high energies to these samples. The authors introduce the Energy‑based Generative Adversarial Network (EBGAN), which treats the discriminator as an energy function assigning low energies near the data manifold and higher energies elsewhere. By modeling the discriminator as an energy function, the framework supports diverse architectures and loss functions; the authors instantiate it with an auto‑encoder whose reconstruction error serves as the energy. EBGAN demonstrates more stable training than conventional GANs and enables a single‑scale architecture to generate high‑resolution images.
We introduce the "Energy-based Generative Adversarial Network" model (EBGAN) which views the discriminator as an energy function that attributes low energies to the regions near the data manifold and higher energies to other regions. Similar to the probabilistic GANs, a generator is seen as being 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, we show one instantiation of EBGAN framework as using an auto-encoder architecture, with the energy being the reconstruction error, in place of the discriminator. 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.
19
Adam: A Method for Stochastic Optimization
Diederik P. Kingma, Jimmy Ba · UvA-DARE (University of Amsterdam) · 2014 · 84.8K citations · Full text
ImageNet Large Scale Visual Recognition Challenge
Olga Russakovsky, Jia Deng, Hao Su et al. · International Journal of Computer Vision · 2015 · 40.2K citations
Deep Learning Face Attributes in the Wild
Ziwei Liu, Ping Luo, Xiaogang Wang et al. · 2015 · 7.7K citations