Note on the geodesic Monte Carlo
share
Geodesic Monte Carlo (gMC) comprises a powerful class of algorithms for Bayesian inference on non-Euclidean manifolds. The original gMC algorithm was cleverly derived in terms of its progenitor, the Riemannian manifold Hamiltonian Monte Carlo (RMHMC). Here, it is shown that an alternative, conceptually simpler derivation is available which clarifies the algorithm when applied to manifolds embedded in Euclidean space.
READ FULL TEXT