An introduction to sampling via measure transport

Abstract

We present the fundamentals of a measure transport approach to sampling. The idea is to construct a deterministic coupling---i.e., a transport map---between a complex "target" probability measure of interest and a simpler reference measure. Given a transport map, one can generate arbitrarily many independent and unweighted samples from the target simply by pushing forward reference samples through the map. We consider two different and complementary scenarios: first, when only evaluations of the unnormalized target density are available, and second, when the target distribution is known only through a finite collection of samples. We show that in both settings the desired transports can be characterized as the solutions of variational problems. We then address practical issues associated with the optimization--based construction of transports: choosing finite-dimensional parameterizations of the map, enforcing monotonicity, quantifying the error of approximate transports, and refining approximate transports by enriching the corresponding approximation spaces. Approximate transports can also be used to "Gaussianize" complex distributions and thus precondition conventional asymptotically exact sampling schemes. We place the measure transport approach in broader context, describing connections with other optimization--based samplers, with inference and density estimation schemes using optimal transport, and with alternative transformation--based approaches to simulation. We also sketch current work aimed at the construction of transport maps in high dimensions, exploiting essential features of the target distribution (e.g., conditional independence, low-rank structure). The approaches and algorithms presented here have direct applications to Bayesian computation and to broader problems of stochastic simulation.

To appear in Handbook of Uncertainty Quantification; R. Ghanem, D. Higdon, and H. Owhadi, editors; Springer, 2016

References in corpus (2)

• Inferring Parameters and Structure of Latent Variable Models by Variational Bayes
• Iterative Gaussianization: from ICA to Random Rotations

Cited by in corpus (32)

• Diffusion Models in Vision: A Survey
• Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
• Seismic tomography using variational inference methods
• Compositionally-Warped Gaussian Processes
• Hessian-based adaptive sparse quadrature for infinite-dimensional Bayesian inverse problems
• A Stein variational Newton method
• On the representation and learning of monotone triangular transport maps
• Deep composition of tensor-trains using squared inverse Rosenblatt transports
• A comparison of nonlinear extensions to the ensemble Kalman filter: Gaussian Anamorphosis and Two-Step Ensemble Filters
• Registration-based model reduction of parameterized two-dimensional conservation laws
• Preconditioned training of normalizing flows for variational inference in inverse problems
• Data-driven low-fidelity models for multi-fidelity Monte Carlo sampling in plasma micro-turbulence analysis
• Bayesian model inversion using stochastic spectral embedding
• Projected Stein Variational Gradient Descent
• Greedy inference with structure-exploiting lazy maps
• Inference via low-dimensional couplings
• Beyond normality: Learning sparse probabilistic graphical models in the non-Gaussian setting
• A low-rank ensemble Kalman filter for elliptic observations
• Multilevel adaptive sparse Leja approximations for Bayesian inverse problems
• Diffeomorphic random sampling using optimal information transport
• Efficient approximation of high-dimensional exponentials by tensornetworks
• Registration-based model reduction in complex two-dimensional geometries
• Scalable optimization-based sampling on function space
• Energy-Based Coarse-Graining in Molecular Dynamics: A Flow-Based Framework without Data
• Posterior Integration on a Riemannian Manifold
• Calibrated Multivariate Regression with Localized PIT Mappings
• Bayesian inverse problems with $l_1$ priors: a Randomize-then-Optimize approach
• Sampling in Combinatorial Spaces with SurVAE Flow Augmented MCMC
• Density Estimation via Measure Transport: Outlook for Applications in the Biological Sciences
• Physics Informed Convex Artificial Neural Networks (PICANNs) for Optimal Transport based Density Estimation
• hIPPYlib-MUQ: A Bayesian Inference Software Framework for Integration of Data with Complex Predictive Models under Uncertainty
• Generative multi-scale modeling via spatial autoregressive transport maps