Graphical Generative Adversarial Networks

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.

Cited by in corpus (10)

• Deep Structural Causal Models for Tractable Counterfactual Inference
• MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering
• Triple Generative Adversarial Networks
• Latent Dirichlet Allocation in Generative Adversarial Networks
• Bi-level Score Matching for Learning Energy-based Latent Variable Models
• Region-based Energy Neural Network for Approximate Inference
• Adversarially-learned Inference via an Ensemble of Discrete Undirected Graphical Models
• You Never Cluster Alone
• On the Generative Utility of Cyclic Conditionals
• Lightweight Data Fusion with Conjugate Mappings