A Parallel Algorithm for Data Clustering

Please enable javascript in your browser.

Ashkboos, Saleh | 2019

452 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 52783 (02)
University: Sharif University of Technology
Department: Mathematical Sciences
Advisor(s): Daneshgar, Amir; Javadi Jurtani, Ramin
Abstract:
Data clustering, is one of the most important problems in computer science. It has countless applica-tions in many different topics like machine learning, image processing, computer vision and bioinfor-matics. However, current algorithms for this problem do not have high accuracy and speed. In this thesis, we have proposed a parallel algorithm for data clustering based on the solution of Isoperimetric problem on trees. In addition, we have implemented the algorithm on GPU as a powerful, cheap and mostly available hardware. Finally, by doing different practical experiments, we have shown that the algorithm has more than 20x speed-up over the sequential algorithm without changing in the accuracy
Keywords:
Data Clustering ; Graph Cut ; Parallel Algorithm ; Isoperimetric Number

No TOC