Distributed Evidence Propagation in Junction Trees

Yinglong Xia, Viktor K. Prasanna
2010 2010 22nd International Symposium on Computer Architecture and High Performance Computing  
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 subtrees, and then perform evidence propagation in the subtrees in parallel. The partially updated subtrees are merged after evidence collection. In addition, we propose a technique to explore tradeoff between overhead due to startup latency of message
more » ... y of message passing and bandwidth utilization efficiency. We implemented the proposed method on state-ofthe-art clusters using MPI. Experimental results show that the proposed method exhibits superior performance compared with the baseline methods.
doi:10.1109/sbac-pad.2010.25 dblp:conf/sbac-pad/XiaP10 fatcat:k2y3e7keg5dntblfzuyohg4afy