Published on Tue Apr 10 2018

Graphical Generative Adversarial Networks

Chongxuan Li, Max Welling, Jun Zhu, Bo Zhang

We propose Graphical Generative Adversarial Networks (Graphical-GAN) to model structured data. Graphicals-GAN conjoins the power of Bayesian networks on representing the dependency structures among random variables and of generative adversarial networks on learning expressive dependency functions.

0
0
0
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.

Tue Oct 09 2018
Machine Learning
Generalized Latent Variable Recovery for Generative Adversarial Networks
Generator of a Generative Adversarial Network (GAN) is trained to transform latent vectors drawn from a prior distribution into realistic looking photos. These latent vectors have been shown to encode information about the corresponding images.
0
0
0
Sat Mar 16 2019
Machine Learning
Generative Adversarial Networks: recent developments
In traditional generative modeling, good data representation is very often a substrate for a good machine learning model. With the invention of Generative Adversarial Networks (GANs), we are now able to adversarially learn good mappings from a simple prior distribution to a target data distribution.
0
0
0
Wed May 27 2020
Machine Learning
Generative Adversarial Networks (GANs): An Overview of Theoretical Model, Evaluation Metrics, and Recent Developments
Generative Adversarial Network (GAN) is an effective method to address this problem. GANs provide an appropriate way to learn deep representations without widespread use of labeled training data.
0
0
0
Mon Mar 12 2018
Neural Networks
Learning the Base Distribution in Implicit Generative Models
We argue that learning a complicated distribution over the latent space of an auto-encoder enables more accurate modeling of complicated data. We propose a two stage optimization procedure which maximizes an approximate implicit density model. We also show that our approach is amenable to learning a sequential data model.
0
0
0
Wed Apr 08 2020
Machine Learning
Mixture Density Conditional Generative Adversarial Network Models (MD-CGAN)
0
0
0
Fri Jan 04 2019
Machine Learning
Adaptive Density Estimation for Generative Models
Unsupervised learning of generative models has seen tremendous progress. GANs have dramatically improved sample quality, but suffer from two drawbacks. These shortcomings can be addressed by training generative latent variable models in ahybrid adversarial-likelihood manner.
0
0
0