Advertisement
Published in:
2018 | OriginalPaper | Chapter
Improved Implementation of Expectation Maximization Algorithm on Graphic Processing Unit
Authors : Si-Yuan Jing, Rui Sun, Chun-Ming Xie, Peng Jin, Yi Liu, Cai-Ming Liu
Published in: Chinese Lexical Semantics
Publisher: Springer International Publishing
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Abstract
In our previous work, an efficient implementation of Expectation-Maximization (EM) algorithm using CUDA has been proposed for high-speed word alignment. The proposed algorithm can gain a 16.8-fold speedup compared to a multi-thread algorithm and a 234.7-fold speedup compared to a sequential algorithm on a modern graphic processing unit (GPU). In this paper, we try to improve the algorithm to achieve better performance. Through analysis of the previous algorithm, we find that two places in “E” step (expectation calculation) are unreasonably designed. An improved CUDA implementation of the EM algorithm is proposed in this paper. Experimental results show that the new algorithm can improve the speed of expectation calculation by 29.4%.