The following information was submitted:
Transactions: INTERNATIONAL JOURNAL of MATHEMATICS AND COMPUTERS IN SIMULATION
Transactions ID Number: 19-600
Full Name: Kittisak Kerdprasop
Position: Associate Professor
Age: ON
Sex: Male
Address: Computer Engineering, Suranaree University of Technology
Country: THAILAND
Tel: +66-44-224349
Tel prefix:
Fax: +66-44-224602
E-mail address: KittisakThailand@gmail.com
Other E-mails: kerdpras@sut.ac.th,profKittisak@gmail.com
Title of the Paper: A lightweight method to parallel k-means clustering
Authors as they appear in the Paper: Kittisak Kerdprasop and Nittaya Kerdprasop
Email addresses of all the authors: KittisakThailand@gmail.com,nittaya@sut.ac.th
Number of paper pages: 10
Abstract: Traditional k-means clustering iteratively performs two major steps: data assignment and calculating the relocation of mean points. The data assignment step sends each data point to a cluster with closest mean, or centroid. Normally, the measure of closeness is the Euclidean distance. On clustering large datasets, the k-means method spends most of its execution time on computing distances between all data points and existing centroids. It is obvious that distance computation of one data point is irrelevant to others. Therefore, data parallelism can be achieved in this case and it is the main focus of this paper. We propose the parallel method as well as its approximation scheme to the k-means clustering. The parallelism is implemented through the message passing model using a concurrent functional language, Erlang. The experimental results show the speedup in computation of parallel k-means. The clustering results of an approximated parallel method are impressive w!
hen taking into account its fast running time.
Keywords: Parallel k-means, Lightweight process, Message passing, Concurrent functional program, Erlang
EXTENSION of the file: .pdf
Special (Invited) Session: Parallelization of k-means clustering on multi-core processors
Organizer of the Session: 635-482
How Did you learn about congress: Professor Junping Sun, jps@nova.edu
IP ADDRESS: 203.158.4.226
