Inference Networks for Sequential Monte Carlo in Graphical Models
arXiv:1602.06701
Abstract
We introduce a new approach for amortizing inference in directed graphical models by learning heuristic approximations to stochastic inverses, designed specifically for use as proposal distributions in sequential Monte Carlo methods. We describe a procedure for constructing and learning a structured neural network which represents an inverse factorization of the graphical model, resulting in a conditional density estimator that takes as input particular values of the observed random variables, and returns an approximation to the distribution of the latent variables. This recognition model can be learned offline, independent from any particular dataset, prior to performing inference. The output of these networks can be used as automatically-learned high-quality proposal distributions to accelerate sequential Monte Carlo across a diverse range of problem settings.
10 pages. Updated from version at ICML 2016; includes code at http://github.com/tbrx/compiled-inference
References in corpus (3)
Cited by in corpus (25)
- The frontier of simulation-based inference
- Masked Autoregressive Flow for Density Estimation
- Mining gold from implicit models to improve likelihood-free inference
- Learning to Draw Samples: With Application to Amortized MLE for Generative Adversarial Learning
- Mining for Dark Matter Substructure: Inferring subhalo population properties from strong lenses with machine learning
- Variational Sequential Monte Carlo
- Deep Amortized Inference for Probabilistic Programs
- ZhuSuan: A Library for Bayesian Deep Learning
- Learning to Draw Samples with Amortized Stein Variational Gradient Descent
- Tighter Variational Bounds are Not Necessarily Better
- Effective LHC measurements with matrix elements and machine learning
- Revisiting Reweighted Wake-Sleep for Models with Stochastic Control Flow
- Likelihood-free inference with an improved cross-entropy estimator
- Some models are useful, but how do we know which ones? Towards a unified Bayesian model taxonomy
- Learning to superoptimize programs
- Constructing Deep Neural Networks by Bayesian Network Structure Learning
- Comparison of Affine and Rational Quadratic Spline Coupling and Autoregressive Flows through Robust Statistical Tests
- Two Methods For Wild Variational Inference
- Faithful Inversion of Generative Models for Effective Amortized Inference
- Using probabilistic programs as proposals
- The divide-and-conquer sequential Monte Carlo algorithm: theoretical properties and limit theorems
- Filtering Variational Objectives
- Iterative Refinement of the Approximate Posterior for Directed Belief Networks
- Amortized Bayesian inference for clustering models
- Neural Particle Smoothing for Sampling from Conditional Sequence Models