Automatically Marginalized MCMC in Probabilistic Programming
This work addresses a bottleneck for users of probabilistic programming languages by simplifying inference in hierarchical models, though it is incremental as it builds on existing HMC and marginalization techniques.
The paper tackled the challenge of sampling from complex Bayesian models in probabilistic programming by introducing automatic marginalization within Hamiltonian Monte Carlo, which substantially improved sampling efficiency for real-world hierarchical models.
Hamiltonian Monte Carlo (HMC) is a powerful algorithm to sample latent variables from Bayesian models. The advent of probabilistic programming languages (PPLs) frees users from writing inference algorithms and lets users focus on modeling. However, many models are difficult for HMC to solve directly, and often require tricks like model reparameterization. We are motivated by the fact that many of those models could be simplified by marginalization. We propose to use automatic marginalization as part of the sampling process using HMC in a graphical model extracted from a PPL, which substantially improves sampling from real-world hierarchical models.