Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks

Abstract

Variational Autoencoders (VAEs) are expressive latent variable models that can be used to learn complex probability distributions from training data. However, the quality of the resulting model crucially relies on the expressiveness of the inference model. We introduce Adversarial Variational Bayes (AVB), a technique for training Variational Autoencoders with arbitrarily expressive inference models. We achieve this by introducing an auxiliary discriminative network that allows to rephrase the maximum-likelihood-problem as a two-player game, hence establishing a principled connection between VAEs and Generative Adversarial Networks (GANs). We show that in the nonparametric limit our method yields an exact maximum-likelihood assignment for the parameters of the generative model, as well as the exact posterior distribution over the latent variables given an observation. Contrary to competing approaches which combine VAEs with GANs, our approach has a clear theoretical justification, retains most advantages of standard Variational Autoencoders and is easy to implement.

References in corpus (4)

• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space
• Automatic Variational Inference in Stan
• Improved generator objectives for GANs

Cited by in corpus (57)

• Disentangling by Factorising
• Learning on Attribute-Missing Graphs
• A survey of synthetic data augmentation methods in computer vision
• Adversarial training with cycle consistency for unsupervised super-resolution in endomicroscopy
• Lifelong Teacher-Student Network Learning
• Distribution Matching in Variational Inference
• An advanced hybrid deep adversarial autoencoder for parameterized nonlinear fluid flow modelling
• Prescribed Generative Adversarial Networks
• A survey on Adversarial Recommender Systems: from Attack/Defense strategies to Generative Adversarial Networks
• Advances in Variational Inference
• Bayesian graph convolutional neural networks via tempered MCMC
• DP-Image: Differential Privacy for Image Data in Feature Space
• Acoustic Leak Detection in Water Networks
• Feature-metric Loss for Self-supervised Learning of Depth and Egomotion
• Biased Mixtures Of Experts: Enabling Computer Vision Inference Under Data Transfer Limitations
• Hierarchical Patch VAE-GAN: Generating Diverse Videos from a Single Sample
• On the Necessity and Effectiveness of Learning the Prior of Variational Auto-Encoder
• A Contrastive Learning Approach for Training Variational Autoencoder Priors
• Learning latent representations across multiple data domains using Lifelong VAEGAN
• Inference over radiative transfer models using variational and expectation maximization methods
• Conditional Adversarial Generative Flow for Controllable Image Synthesis
• Variational Autoencoder with Implicit Optimal Priors
• Factor Analysis, Probabilistic Principal Component Analysis, Variational Inference, and Variational Autoencoder: Tutorial and Survey
• Implicit Deep Latent Variable Models for Text Generation
• Bidirectional Generative Modeling Using Adversarial Gradient Estimation
• Posterior Meta-Replay for Continual Learning
• Adversarial and Contrastive Variational Autoencoder for Sequential Recommendation
• APo-VAE: Text Generation in Hyperbolic Space
• Variational Selective Autoencoder: Learning from Partially-Observed Heterogeneous Data
• KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
• Variational Inference with Holder Bounds
• Importance Weighted Hierarchical Variational Inference
• Inferential Wasserstein Generative Adversarial Networks
• Regularized Inverse Reinforcement Learning
• Generative Adversarial Networks and Adversarial Autoencoders: Tutorial and Survey
• Learning More with Less: Conditional PGGAN-based Data Augmentation for Brain Metastases Detection Using Highly-Rough Annotation on MR Images
• Mind the Gap when Conditioning Amortised Inference in Sequential Latent-Variable Models
• Copula-like Variational Inference
• Regularized Sequential Latent Variable Models with Adversarial Neural Networks
• Deep Image Synthesis from Intuitive User Input: A Review and Perspectives
• Reliable Estimation of KL Divergence using a Discriminator in Reproducing Kernel Hilbert Space
• Scalable Approximate Inference and Some Applications
• NeurInt : Learning to Interpolate through Neural ODEs
• End-To-End Dilated Variational Autoencoder with Bottleneck Discriminative Loss for Sound Morphing -- A Preliminary Study
• Dual Adversarial Variational Embedding for Robust Recommendation
• Deep Dive into Semi-Supervised ELBO for Improving Classification Performance
• MCMC-Interactive Variational Inference
• AriEL: volume coding for sentence generation
• Stacked Wasserstein Autoencoder
• Analysis of Discriminator in RKHS Function Space for Kullback-Leibler Divergence Estimation
• Deep Automodulators
• Multi-class Novelty Detection Using Mix-up Technique
• Variational Hetero-Encoder Randomized GANs for Joint Image-Text Modeling
• Differential Similarity in Higher Dimensional Spaces: Theory and Applications
• Disentangling Latent Space for VAE by Label Relevant/Irrelevant Dimensions
• Coulomb Autoencoders
• Adversarial Code Learning for Image Generation