Bayesian structure learning using dynamic programming and MCMC

Daniel Eaton, Kevin Murphy

MCMC methods for sampling from the space of DAGs can mix poorly due to the local nature of the proposals that are commonly used. It has been shown that sampling from the space of n…