MCMC using Hamiltonian dynamics — leapfrog integration proposes distant moves with high acceptance
Sample Space (q)
Phase Space (q₁, p₁)
Ready — click Run HMC
HMC augments state q with momentum p, then simulates Hamiltonian H(q,p) = U(q) + K(p) via leapfrog.
The proposal travels along constant-energy contours — exploring efficiently without random-walk diffusion.
Orange trajectory = current leapfrog path. Blue dots = accepted samples. Red = rejected.