DocumentCode :
3224611
Title :
Distributed Evidence Propagation in Junction Trees
Author :
Xia, Yinglong ; Prasanna, Viktor K.
Author_Institution :
Comput. Sci. Dept., Univ. of Southern California, Los Angeles, CA, USA
fYear :
2010
fDate :
27-30 Oct. 2010
Firstpage :
143
Lastpage :
150
Abstract :
Evidence propagation is a major step in exact inference, a key problem in exploring probabilistic graphical models. In this paper, we propose a novel approach for evidence propagation on clusters. We decompose a junction tree into a set of sub trees, and then perform evidence propagation in the sub trees in parallel. The partially updated sub trees are merged after evidence collection. In addition, we propose a technique to explore tradeoff between overhead due to startup latency of message passing and bandwidth utilization efficiency. We implemented the proposed method on state-of-the-art clusters using MPI. Experimental results show that the proposed method exhibits superior performance compared with the baseline methods.
Keywords :
message passing; pattern clustering; MPI; bandwidth utilization; distributed evidence propagation; junction tree; message passing; Bandwidth; Bayesian methods; Junctions; Merging; Particle separators; Program processors; Silicon; cluster; exact inference; junction tree; parallel computing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Architecture and High Performance Computing (SBAC-PAD), 2010 22nd International Symposium on
Conference_Location :
Petropolis
ISSN :
1550-6533
Print_ISBN :
978-1-4244-8287-0
Electronic_ISBN :
1550-6533
Type :
conf
DOI :
10.1109/SBAC-PAD.2010.25
Filename :
5644957
Link To Document :
بازگشت