Markov Chain Monte Carlo and Variational Inference: Bridging the Gap

Abstract

Recent advances in stochastic gradient variational inference have made it possible to perform variational Bayesian inference with posterior approximations containing auxiliary random variables. This enables us to explore a new synthesis of variational inference and Monte Carlo methods where we incorporate one or more steps of MCMC into our variational approximation. By doing so we obtain a rich class of inference algorithms bridging the gap between variational methods and MCMC, and offering the best of both worlds: fast posterior approximation through the maximization of an explicit objective, with the option of trading off additional computation for additional accuracy. We describe the theoretical foundations that make this possible and show some promising first results.

References in corpus (4)

• Theano: new features and speed improvements
• DRAW: A Recurrent Neural Network For Image Generation
• Variational Bayesian Inference with Stochastic Search
• How to view an MCMC simulation as a permutation, with applications to parallel simulation and improved importance sampling

Cited by in corpus (49)

• Professor Forcing: A New Algorithm for Training Recurrent Networks
• Improving Variational Auto-Encoders using Householder Flow
• NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport
• On denoising autoencoders trained to minimise binary cross-entropy
• Normalizing Flows on Riemannian Manifolds
• The Free Energy Principle for Perception and Action: A Deep Learning Perspective
• Stochastic Normalizing Flows
• Equivariant Hamiltonian Flows
• ZhuSuan: A Library for Bayesian Deep Learning
• An introduction to variational inference in Geophysical inverse problems
• On the challenges of learning with inference networks on sparse, high-dimensional data
• Learning to Draw Samples with Amortized Stein Variational Gradient Descent
• Constructing dynamic residential energy lifestyles using Latent Dirichlet Allocation
• Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models
• Designing Perceptual Puzzles by Differentiating Probabilistic Programs
• Learning Model Reparametrizations: Implicit Variational Inference by Fitting MCMC distributions
• Understanding the (un)interpretability of natural image distributions using generative models
• On the Necessity and Effectiveness of Learning the Prior of Variational Auto-Encoder
• Probabilistic Programming with Programmable Variational Inference
• Deep Variational Inference Without Pixel-Wise Reconstruction
• Deep Markov Chain Monte Carlo
• MetFlow: A New Efficient Method for Bridging the Gap between Markov Chain Monte Carlo and Variational Inference
• Data Augmentation for Enhancing EEG-based Emotion Recognition with Deep Generative Models
• An Uncertain Future: Forecasting from Static Images using Variational Autoencoders
• CDVAE: Co-embedding Deep Variational Auto Encoder for Conditional Variational Generation
• Joint Stochastic Approximation and Its Application to Learning Discrete Latent Variable Models
• Combining Model and Parameter Uncertainty in Bayesian Neural Networks
• Accelerating MCMC algorithms through Bayesian Deep Networks
• Importance Weighted Hierarchical Variational Inference
• All in the Exponential Family: Bregman Duality in Thermodynamic Variational Inference
• Neural Variational Inference and Learning in Undirected Graphical Models
• Energy-Inspired Models: Learning with Sampler-Induced Distributions
• Measure Transport with Kernel Stein Discrepancy
• MCMC Variational Inference via Uncorrected Hamiltonian Annealing
• Differentiable Annealed Importance Sampling and the Perils of Gradient Noise
• Cauchy-Schwarz Regularized Autoencoder
• Generative networks as inverse problems with fractional wavelet scattering networks
• Soft-Deep Boltzmann Machines
• Learning to Decipher Hate Symbols
• Reparameterized Sampling for Generative Adversarial Networks
• Decoupled Variational Embedding for Signed Directed Networks
• Enhanced Variational Inference with Dyadic Transformation
• Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling
• A Divergence Bound for Hybrids of MCMC and Variational Inference and an Application to Langevin Dynamics and SGVI
• A Hybrid Approximation to the Marginal Likelihood
• Self-Reflective Variational Autoencoder
• Quantized Variational Inference
• Coarse Grained Exponential Variational Autoencoders
• Contributions to Large Scale Bayesian Inference and Adversarial Machine Learning