Understanding the Hamiltonian Monte Carlo through its Physics Fundamentals and Examples

The Hamiltonian Monte Carlo (HMC) algorithm is a powerful Markov Chain Monte Carlo (MCMC) method that uses Hamiltonian dynamics to generate samples from a target distribution.

To fully exploit its potential, we must understand how Hamiltonian dynamics work and why they can be used in a MCMC algorithm.

This work elucidates the Monte Carlo Hamiltonian, providing comprehensive explanations of the underlying physical concepts.

It is intended for readers with a solid foundation in mathematics who may lack familiarity with specific physical concepts, such as those related to Hamiltonian dynamics.

Additionally, we provide Python code for the HMC algorithm, examples and comparisons with the Random Walk Metropolis-Hastings (RWMH) algorithm, alongside an exploration of modern variants such as the No-U-Turn Sampler (NUTS), the Dual Averaging (DA) scheme and the repelling-attracting HMC (raHMC), to highlight HMC's strengths and weaknesses when applied to Bayesian Inference.