Stein Variational Gradient Descent as Gradient Flow

Abstract

Stein variational gradient descent (SVGD) is a deterministic sampling algorithm that iteratively transports a set of particles to approximate given distributions, based on an efficient gradient-based update that guarantees to optimally decrease the KL divergence within a function space. This paper develops the first theoretical analysis on SVGD, discussing its weak convergence properties and showing that its asymptotic behavior is captured by a gradient flow of the KL divergence functional under a new metric structure induced by Stein operator. We also provide a number of results on Stein operator and Stein's identity using the notion of weak derivative, including a new proof of the distinguishability of Stein discrepancy under weak conditions.

Cited by in corpus (58)

• Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification
• Quantifying model form uncertainty in Reynolds-averaged turbulence models with Bayesian deep neural networks
• Interacting particle solutions of Fokker-Planck equations through gradient-log-density estimation
• Maximum Mean Discrepancy Gradient Flow
• Stein Variational Gradient Descent With Matrix-Valued Kernels
• A Stein variational Newton method
• On the geometry of Stein variational gradient descent
• A stochastic version of Stein Variational Gradient Descent for efficient sampling
• An introduction to variational inference in Geophysical inverse problems
• A Non-Asymptotic Analysis for Stein Variational Gradient Descent
• Sampling as optimization in the space of measures: The Langevin dynamics as a composite optimization problem
• Stein Variational Gradient Descent Without Gradient
• Stochastic Gradient MCMC with Repulsive Forces
• Accelerating Langevin Sampling with Birth-death
• Stein variational gradient descent with local approximations
• Particle-based Energetic Variational Inference
• Repulsive Deep Ensembles are Bayesian
• Greedy inference with structure-exploiting lazy maps
• Particle Optimization in Stochastic Gradient MCMC
• Stochastic Stein Discrepancies
• Information Newton's flow: second-order optimization method in probability space
• Scaling limit of the Stein variational gradient descent: the mean field regime
• Wasserstein variational gradient descent: From semi-discrete optimal transport to ensemble variational inference
• Straight-Through Estimator as Projected Wasserstein Gradient Flow
• Stein Variational Online Changepoint Detection with Applications to Hawkes Processes and Neural Networks
• Parameter Estimation for the McKean-Vlasov Stochastic Differential Equation
• On Stein Variational Neural Network Ensembles
• Stein Self-Repulsive Dynamics: Benefits From Past Samples
• Kernel Stein Generative Modeling
• Applying SVGD to Bayesian Neural Networks for Cyclical Time-Series Prediction and Inference
• New particle representations for ergodic McKean-Vlasov SDEs
• Deep Generative Learning via Variational Gradient Flow
• Sinkhorn Barycenter via Functional Gradient Descent
• KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
• Stein Variational Inference for Discrete Distributions
• Stein Variational Gaussian Processes
• Refining Deep Generative Models via Discriminator Gradient Flow
• The Wasserstein Proximal Gradient Algorithm
• Variational Transport: A Convergent Particle-BasedAlgorithm for Distributional Optimization
• Loss function based second-order Jensen inequality and its application to particle variational inference
• Sampling with Mirrored Stein Operators
• Generative Learning With Euler Particle Transport
• Understanding MCMC Dynamics as Flows on the Wasserstein Space
• The equivalence between Stein variational gradient descent and black-box variational inference
• Stein variational reduced basis Bayesian inversion
• Bayesian posterior approximation via greedy particle optimization
• Hessian transport gradient flows
• Scalable Approximate Inference and Some Applications
• Doubly Robust Stein-Kernelized Monte Carlo Estimator: Simultaneous Bias-Variance Reduction and Supercanonical Convergence
• Stabilizing Training of Generative Adversarial Nets via Langevin Stein Variational Gradient Descent
• A Variational View on Bootstrap Ensembles as Bayesian Inference
• Accelerated Information Gradient flow
• Neural Variational Gradient Descent
• Contributions to Large Scale Bayesian Inference and Adversarial Machine Learning
• Generative Particle Variational Inference via Estimation of Functional Gradients
• Relative Entropy Gradient Sampler for Unnormalized Distributions
• Multilevel Stein variational gradient descent with applications to Bayesian inverse problems
• A Wasserstein Minimum Velocity Approach to Learning Unnormalized Models