Learning to Draw Samples with Amortized Stein Variational Gradient Descent

Abstract

We propose a simple algorithm to train stochastic neural networks to draw samples from given target distributions for probabilistic inference. Our method is based on iteratively adjusting the neural network parameters so that the output changes along a Stein variational gradient direction (Liu & Wang, 2016) that maximally decreases the KL divergence with the target distribution. Our method works for any target distribution specified by their unnormalized density function, and can train any black-box architectures that are differentiable in terms of the parameters we want to adapt. We demonstrate our method with a number of applications, including variational autoencoder (VAE) with expressive encoders to model complex latent space structures, and hyper-parameter learning of MCMC samplers that allows Bayesian inference to adaptively improve itself when seeing more data.

Accepted by UAI 2017

References in corpus (8)

• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
• Learning in Implicit Generative Models
• Improving Variational Inference with Inverse Autoregressive Flow
• A Variational Analysis of Stochastic Gradient Algorithms
• Inference Networks for Sequential Monte Carlo in Graphical Models
• Approximate Inference with Amortised MCMC
• Early Stopping is Nonparametric Variational Inference

Cited by in corpus (7)

• A Non-Asymptotic Analysis for Stein Variational Gradient Descent
• Metrizing Weak Convergence with Maximum Mean Discrepancies
• Sliced Kernelized Stein Discrepancy
• Tractable Regularization of Probabilistic Circuits
• Implicit Generative Modeling for Efficient Exploration
• Variationally Inferred Sampling Through a Refined Bound for Probabilistic Programs
• Contributions to Large Scale Bayesian Inference and Adversarial Machine Learning