Title :
Parallel sampling in Bayesian networks
Author :
Rego, Vernon ; Schulz, Andrew
Author_Institution :
Dept. of Comput. Sci., Purdue Univ., West Lafayette, IN, USA
Abstract :
An easily parallelized version of the Pearl sequential algorithm is presented, along with experimental results utilizing an Ncube/1, hypercube and a Sequent shared memory multiprocessor. Pearl´s concurrent simulation algorithm is briefly reviewed, and three modifications to this algorithm are proposed. These modifications, known as stack, phase, and parallel simulations, show considerable speedup in experiments performed on the hypercube
Keywords :
Bayes methods; hypercube networks; multiprocessing systems; parallel algorithms; Bayesian networks; Ncube/1; Pearl sequential algorithm; Sequent shared memory multiprocessor; concurrent simulation algorithm; hypercube; parallel sampling; Bayesian methods; Computer networks; Concurrent computing; Inference algorithms; Intelligent networks; Logic; Parallel processing; Random variables; Sampling methods; Uncertainty;
Conference_Titel :
Tools for Artificial Intelligence, 1991. TAI '91., Third International Conference on
Conference_Location :
San Jose, CA
Print_ISBN :
0-8186-2300-4
DOI :
10.1109/TAI.1991.167077
