trends.barnas.net

method · first seen 12 Oct 2026

Hamiltonian Monte Carlo

Also: HMC, hybrid Monte Carlo

Markov chain Monte Carlo sampling method that uses Hamiltonian dynamics to propose distant, high-acceptance moves.

Hamiltonian Monte Carlo is an MCMC method that simulates Hamiltonian dynamics with auxiliary momentum variables to make long-range proposals with high acceptance rates. It dates to the 1987 lattice QCD work by Duane and colleagues and is the default sampler in tools like Stan through the No-U-Turn variant. The papers in this window are incremental. They cover locally adaptive step sizes for NUTS, federated HMC with lower communication cost, mixing-time theory for kinetic Langevin, a survey of HMC for hierarchical models, unadjusted samplers for quantum Monte Carlo, and applied Bayesian studies that use HMC as a standard inference tool.

HMC remains core infrastructure for Bayesian inference and lattice physics, so variants that improve step-size adaptation or scale to federated settings can affect a wide range of applied work. The small rise in mentions here reflects normal research output, not new adoption.

Analyst note (established), 18 Nov 2026.

Mentions per week

0319 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

No-U-Turn samplerconformal predictionvariational inferenceBayesian predictive inferencediffusion modelsBayesian hierarchical regressionGaussian process classificationrandomly pivoted Choleskyuncertainty quantification under distribution shiftunadjusted Langevin-type sampling