Parallel Exact Inference.

摘要: In this paper, we present complete message-passing implementation that shows scalable performance while performing exact inference on arbitrary Bayesian networks. Our work is based a parallel version of the classical technique converting network to junction tree before computing inference. We propose algorithm for constructing potential tables and explore parallelism rerooting multiple evidence propagation. also uses pointer jumping over tree. For an with n vertices using p processors, show execution time O(nk m + nw (nw wN log rwwN rwN N)/p), where w clique width, r number states random variables, k maximum node degree in network, km moralized graph N cliques shown be 1 ≤ moralization identification, construction, table have implemented MPI state-of-the-art clusters our experiments performance.