Abstract | ||
---|---|---|
In recent years, the advances in microprocessors and high-speed networks are making it possible for scientific applications to run on clustered type of environment. Expectation Maximization (EM) algorithm is a popular iterative method that requires a significant amount of computation and memory in approximating the incomplete data for many real-world problems. In this paper lye propose a sparse matrix compaction technique to speed up the computation by better manipulating the probability matrix. The sparse matrix compaction method is made more efficient by both taking advantage of the geometrical information of the application and removing of indirect addressing overheads that is associated with most of the compressed sparse matrix method. Load balancing is also achieved by a better distribution scheme implemented in the pre-processing phase of the algorithm. By using the proposed method in handling the sparse matrix, our study demonstrates a promising result for the parallelization of the algorithm. |
Year | Venue | Keywords |
---|---|---|
1999 | INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOL VI, PROCEEDINGS | em algorithm,sparse matrix |
Field | DocType | Citations |
Computer science,Expectation–maximization algorithm,Parallel computing,Sparse approximation,Cuthill–McKee algorithm,Compaction,Sparse matrix | Conference | 0 |
PageRank | References | Authors |
0.34 | 1 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Wei-Min Jeng | 1 | 3 | 2.21 |
Shou-hsuan Stephen Huang | 2 | 174 | 59.88 |