Title :
A Parallel Irregular Wavefront Algorithm for Importance Sampling of Probabilistic Networks on GPU
Author :
Yu, Haohai ; van Engelen, R.
Author_Institution :
Dept. of Comput. Sci., Florida State Univ., Tallahassee, FL, USA
Abstract :
Importance sampling is a widely-used method for probabilistic inference with Bayesian probabilistic networks. Importance sampling is relatively easy to parallelize and parallel GPU implementations yield significant speedups over single-CPU implementations. However, because of physical limitations of GPU memory size and bandwidth, the maximum speedups that can be achieved are bounded by the high data transfer requirements of these algorithms. In this paper, we propose and evaluate a new parallel irregular wave front algorithm for importance sampling of probabilistic networks on GPU. Performance results show that the proposed parallel algorithm achieves greater speedups due to the optimal local memory access compared to simple parallel GPU implementations.
Keywords :
computer graphic equipment; coprocessors; importance sampling; inference mechanisms; parallel algorithms; probability; Bayesian probabilistic network; GPU memory size; data transfer requirement; importance sampling; optimal local memory access; parallel GPU implementation; parallel irregular wavefront algorithm; probabilistic inference; single CPU implementation; Bayesian methods; Copper; Graphics processing unit; Instruction sets; Level set; Monte Carlo methods; Probabilistic logic; Bayesian inference; GPGPU parallelization; Importance sampling; Probabilistic graph models;
Conference_Titel :
High Performance Computing and Communications (HPCC), 2011 IEEE 13th International Conference on
Conference_Location :
Banff, AB
Print_ISBN :
978-1-4577-1564-8
Electronic_ISBN :
978-0-7695-4538-7
DOI :
10.1109/HPCC.2011.31
