"Best-of-Many-Samples" Distribution Matching

by   Apratim Bhattacharyya, et al.

Generative Adversarial Networks (GANs) can achieve state-of-the-art sample quality in generative modelling tasks but suffer from the mode collapse problem. Variational Autoencoders (VAE) on the other hand explicitly maximize a reconstruction-based data log-likelihood forcing it to cover all modes, but suffer from poorer sample quality. Recent works have proposed hybrid VAE-GAN frameworks which integrate a GAN-based synthetic likelihood to the VAE objective to address both the mode collapse and sample quality issues, with limited success. This is because the VAE objective forces a trade-off between the data log-likelihood and divergence to the latent prior. The synthetic likelihood ratio term also shows instability during training. We propose a novel objective with a "Best-of-Many-Samples" reconstruction cost and a stable direct estimate of the synthetic likelihood. This enables our hybrid VAE-GAN framework to achieve high data log-likelihood and low divergence to the latent prior at the same time and shows significant improvement over both hybrid VAE-GANS and plain GANs in mode coverage and quality.


page 6

page 8

page 13

page 14


Prescribed Generative Adversarial Networks

Generative adversarial networks (GANs) are a powerful approach to unsupe...

Out-of-Sample Testing for GANs

We propose a new method to evaluate GANs, namely EvalGAN. EvalGAN relies...

Exponentially Tilted Gaussian Prior for Variational Autoencoder

An important propertyfor deep neural networks to possess is the ability ...

Distribution Matching in Variational Inference

The difficulties in matching the latent posterior to the prior, balancin...

A Unified f-divergence Framework Generalizing VAE and GAN

Developing deep generative models that flexibly incorporate diverse meas...

Adversarial Networks and Autoencoders: The Primal-Dual Relationship and Generalization Bounds

Since the introduction of Generative Adversarial Networks (GANs) and Var...

Improved BiGAN training with marginal likelihood equalization

We propose a novel training procedure for improving the performance of g...

Code Repositories


Official implementation of "GAN-Leaks: A Taxonomy of Membership Inference Attacks against Generative Models" (CCS 2020)

view repo