Implementation of K-means on the Cell Broadband Engine Filip imek - PowerPoint PPT Presentation

IBM - CVUT Student Research Projects Implementation of K-means on the Cell Broadband Engine Filip Šimek (simekf1@fel.cvut.cz)

About K-means • a data clustering algorithm • finds clusters of points in space through an iterative process Input data • the more points the more ⇩ ⇩ calculations needed • computaion can be parallelized for better performance Clusterized IBM - CVUT Student Research Projects 2

Cell Broadband Engine • IBM / Sony / Toshiba • Playstation 3 • PowerPC core • 8 Processing Elements • 3.2GHz IBM - CVUT Student Research Projects 3

Project status • algorithm implemented on the PowerPC core of the CBE • created some testing and demo tools • tested on SystemSim CBE simulator • works with multidimensional data IBM - CVUT Student Research Projects 4

Demo application • reducing number of colors of an image Truecolor original 16 colors found using K-means • K-means clusters similar colors and selects a represenatative color for each cluster IBM - CVUT Student Research Projects 5

Future plans • speed up the implementation by using the eight processing cores of the CBE • run K-means more than once on the same data and select the best result • separate learning and classification – adding new points to clusters that are already generated – this will improve execution time of some applications IBM - CVUT Student Research Projects 6