Graphical Generative Adversarial Networks

Authors:
Chongxuan LI Tsinghua University
Max Welling University of Amsterdam / Qualcomm AI Research
Jun Zhu Tsinghua University
Bo Zhang Tsinghua University

Introduction:

The authors propose Graphical Generative Adversarial Networks (Graphical-GAN) to model structured data.

Abstract:

We propose Graphical Generative Adversarial Networks (Graphical-GAN) to model structured data. Graphical-GAN conjoins the power of Bayesian networks on compactly representing the dependency structures among random variables and that of generative adversarial networks on learning expressive dependency functions. We introduce a structured recognition model to infer the posterior distribution of latent variables given observations. We generalize the Expectation Propagation (EP) algorithm to learn the generative model and recognition model jointly. Finally, we present two important instances of Graphical-GAN, i.e. Gaussian Mixture GAN (GMGAN) and State Space GAN (SSGAN), which can successfully learn the discrete and temporal structures on visual datasets, respectively.

You may want to know: