Posterior Integration on a Riemannian Manifold
arXiv:1712.01793
Abstract
The geodesic Markov chain Monte Carlo method and its variants enable computation of integrals with respect to a posterior supported on a manifold. However, for regular integrals, the convergence rate of the ergodic average will be sub-optimal. To fill this gap, this paper extends the efficient posterior integration method of Oates et al. (2017) to the case of a Riemannian manifold. In contrast to the original Euclidean case, no non-trivial boundary conditions are needed for a closed manifold. The method is assessed through simulation and deployed to compute posterior integrals for an Australian Mesozoic paleomagnetic pole model, whose parameters are constrained to lie on the manifold .
arXiv commentThis paper was superseded by arXiv:1810.04946
References in corpus (10)
- Zero-variance principle for Monte Carlo algorithms
- An introduction to sampling via measure transport
- Sparse Deterministic Approximation of Bayesian Inverse Problems
- Measuring Sample Quality with Kernels
- Zero Variance Markov Chain Monte Carlo for Bayesian Estimators
- Embedding and knotting of manifolds in Euclidean spaces
- Spherical Hamiltonian Monte Carlo for Constrained Target Distributions
- A characterization of Sobolev spaces on the sphere and an extension of Stolarsky's invariance principle to arbitrary smoothness
- Score matching estimators for directional distributions
- Riemannian Stein Variational Gradient Descent for Bayesian Inference