Nested Sampling with Constrained Hamiltonian Monte Carlo

A suite of classes facilitating Hamiltonian Monte Carlo (both constrained and unconstrainted) and its use in nested sampling.

baseChain implements the necessary framework for a HMC Morkov chain, including contrainted HMC when sampling from a distribution with bounded support.

chainBundle organizes a container of baseChains, including the functionality to sample from the chains, burn in, and diagnose burn in.

chainNest implements nested sampling, including early termination determination and the calculation of posterior expectations.

An example implementation of a binomial likelihood with beta prior is also included in betaNestedObject and binomialNest. This implemenation also demonstrates the use of a Variational Bayes Gaussian Mixture Model to infer multiple modes in the posterior samples.

Note that a ROOT (http://root.cern.ch/drupal/) installation is required for TRandom3, a Mersenne twistor psuedo-random number generator. If an additional generator is provided then the code would be removed of all dependencies outside of the C++ stdlib (modulo a few mathematical libraries used in gaussMixer).


Generated on Wed Apr 6 18:49:00 2011 by  doxygen 1.5.7