Hybrid unadjusted Langevin methods for high-dimensional latent variable models

TL;DR

This paper introduces a novel Langevin diffusion-based Bayesian method for high-dimensional latent variable models, enabling accurate and computationally efficient estimation in big data contexts without parametric assumptions.

Contribution

It develops a Fisher identity-based Langevin approach that avoids parametric assumptions, improving accuracy and efficiency in large-scale latent variable model estimation.

Findings

Accurately estimates posterior choice probabilities in large datasets

Uses only 2% of the computation time compared to exact MCMC

Demonstrates effectiveness on a million observation dataset

Abstract

The exact estimation of latent variable models with big data is known to be challenging. The latents have to be integrated out numerically, and the dimension of the latent variables increases with the sample size. This paper develops a novel approximate Bayesian method based on the Langevin diffusion process. The method employs the Fisher identity to integrate out the latent variables, which makes it accurate and computationally feasible when applied to big data. In contrast to other approximate estimation methods, it does not require the choice of a parametric distribution for the unknowns, which often leads to inaccuracies. In an empirical discrete choice example with a million observations, the proposed method accurately estimates the posterior choice probabilities using only 2% of the computation time of exact MCMC.