$ ⌘K
// constructs / hamiltonian_monte_carlo

Hamiltonian Monte Carlo (HMC)

id: hamiltonian_monte_carlo

An MCMC algorithm that uses gradient information from the log-posterior to simulate Hamiltonian dynamics, producing long-range proposals with high acceptance rates. Scales substantially better than random-walk Metropolis in moderate-to-high dimensions and is the inference engine underlying Stan and PyMC.

// usage

Used in 1 pax

↑ positive relationship 2 findings