Saved in:
Bibliographic Details
Main Authors: Luu, Son, Xu, Zuheng, Surjanovic, Nikola, Biron-Lattes, Miguel, Campbell, Trevor, Bouchard-Côté, Alexandre
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.03630
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909535968428032
author Luu, Son
Xu, Zuheng
Surjanovic, Nikola
Biron-Lattes, Miguel
Campbell, Trevor
Bouchard-Côté, Alexandre
author_facet Luu, Son
Xu, Zuheng
Surjanovic, Nikola
Biron-Lattes, Miguel
Campbell, Trevor
Bouchard-Côté, Alexandre
contents The Hamiltonian Monte Carlo (HMC) algorithm is often lauded for its ability to effectively sample from high-dimensional distributions. In this paper we challenge the presumed domination of HMC for the Bayesian analysis of GLMs. By utilizing the structure of the compute graph rather than the graphical model, we show a reduction of the time per sweep of a full-scan Gibbs sampler from $O(d^2)$ to $O(d)$, where $d$ is the number of GLM parameters. A simple change to the implementation of the Gibbs sampler allows us to perform Bayesian inference on high-dimensional GLMs that are practically infeasible with traditional Gibbs sampler implementations. We empirically demonstrate a substantial increase in effective sample size per time when comparing our Gibbs algorithms to state-of-the-art HMC algorithms. While Gibbs is superior in terms of dimension scaling, neither Gibbs nor HMC dominate the other: we provide numerical and theoretical evidence that HMC retains an edge in certain circumstances thanks to its advantageous condition number scaling. Interestingly, for GLMs of fixed data size, we observe that increasing dimensionality can stabilize or even decrease condition number, shedding light on the empirical advantage of our efficient Gibbs sampler.
format Preprint
id arxiv_https___arxiv_org_abs_2410_03630
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Is Gibbs sampling faster than Hamiltonian Monte Carlo on GLMs?
Luu, Son
Xu, Zuheng
Surjanovic, Nikola
Biron-Lattes, Miguel
Campbell, Trevor
Bouchard-Côté, Alexandre
Computation
Methodology
The Hamiltonian Monte Carlo (HMC) algorithm is often lauded for its ability to effectively sample from high-dimensional distributions. In this paper we challenge the presumed domination of HMC for the Bayesian analysis of GLMs. By utilizing the structure of the compute graph rather than the graphical model, we show a reduction of the time per sweep of a full-scan Gibbs sampler from $O(d^2)$ to $O(d)$, where $d$ is the number of GLM parameters. A simple change to the implementation of the Gibbs sampler allows us to perform Bayesian inference on high-dimensional GLMs that are practically infeasible with traditional Gibbs sampler implementations. We empirically demonstrate a substantial increase in effective sample size per time when comparing our Gibbs algorithms to state-of-the-art HMC algorithms. While Gibbs is superior in terms of dimension scaling, neither Gibbs nor HMC dominate the other: we provide numerical and theoretical evidence that HMC retains an edge in certain circumstances thanks to its advantageous condition number scaling. Interestingly, for GLMs of fixed data size, we observe that increasing dimensionality can stabilize or even decrease condition number, shedding light on the empirical advantage of our efficient Gibbs sampler.
title Is Gibbs sampling faster than Hamiltonian Monte Carlo on GLMs?
topic Computation
Methodology
url https://arxiv.org/abs/2410.03630