DocumentCode :
3729239
Title :
Scaling GMM Expectation Maximization algorithm using bulk synchronous Parallel approach
Author :
Abhay A. Ratnaparkhi;Emmanuel Pilli;R. C. Joshi
Author_Institution :
Department of Computer Science and Engineering, Graphic Era University, Dehradun, India
fYear :
2015
Firstpage :
558
Lastpage :
562
Abstract :
We have provided a parallel implementation of Gaussian Mixture Model (GMM) Expectation Maximization algorithm using Apache Hama Bulk synchronous Parallel approach. Apache Hama is suitable for iterative, compute intensive tasks. EM is iterative algorithm which converges to local minimum after many iterations. We have provided approach for distributing workload for Expectation and Maximization tasks on cluster nodes in case of big data. The approach is compared with Hadoop MaprRduce and Apache Spark implementations, using different datasets.
Keywords :
"Clustering algorithms","Computational modeling","Sparks","Peer-to-peer computing","Synchronization","Machine learning algorithms","Probability"
Publisher :
ieee
Conference_Titel :
Green Computing and Internet of Things (ICGCIoT), 2015 International Conference on
Type :
conf
DOI :
10.1109/ICGCIoT.2015.7380527
Filename :
7380527
Link To Document :
بازگشت