A Study on the Problem of Maximal Clique Enumeration on GPU
碩士 === 國立臺灣科技大學 === 電子工程系 === 105 === Maximal Clique Enumeration is a graph theory problem, which can find all complete subgraphs in a graph, is suitable for graph analysis such as network structure, graph clustering and community group detection, etc. Bron-Kerbosch algorithm is the most popular alg...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2017
|
Online Access: | http://ndltd.ncl.edu.tw/handle/9374y3 |
id |
ndltd-TW-105NTUS5428162 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-105NTUS54281622019-05-15T23:46:35Z http://ndltd.ncl.edu.tw/handle/9374y3 A Study on the Problem of Maximal Clique Enumeration on GPU 列舉極大完全子圖問題於GPU運算架構之研究 Tsung-Han Wu 吳宗翰 碩士 國立臺灣科技大學 電子工程系 105 Maximal Clique Enumeration is a graph theory problem, which can find all complete subgraphs in a graph, is suitable for graph analysis such as network structure, graph clustering and community group detection, etc. Bron-Kerbosch algorithm is the most popular algorithm to solve this problem, but most of the studies about the Bron-Kerbosch algorithm were in CPU single core execution implementation. It is only a few study on the parallel execution implementation version. Moreover, the general-purpose GPU (GPGPU) computation development is prevalent, which is more efficient about massive computation. In this paper, we proposed a method called GPU-based MCE, which used multiple flags to combine Bron-Kerbosch algorithm three data sets into one data set. Set-manipulation operations in Bron-Kerbosch algorithm were executed based on node-thread one-to-one mappings. We used binary search for the set-manipulation operations to balance the workload in GPU. In addition, we used enhanced atomic operations in the latest NVIDIA CUDA architecture, Pascal, to sum up the pivot score, making the proposed algorithm can get efficient results when executing in parallel using the GPU. Wei-Mei Chen 陳維美 2017 學位論文 ; thesis 54 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺灣科技大學 === 電子工程系 === 105 === Maximal Clique Enumeration is a graph theory problem, which can find all complete subgraphs in a graph, is suitable for graph analysis such as network structure, graph clustering and community group detection, etc. Bron-Kerbosch algorithm is the most popular algorithm to solve this problem, but most of the studies about the Bron-Kerbosch algorithm were in CPU single core execution implementation. It is only a few study on the parallel execution implementation version. Moreover, the general-purpose GPU (GPGPU) computation development is prevalent, which is more efficient about massive computation. In this paper, we proposed a method called GPU-based MCE, which used multiple flags to combine Bron-Kerbosch algorithm three data sets into one data set. Set-manipulation operations in Bron-Kerbosch algorithm were executed based on node-thread one-to-one mappings. We used binary search for the set-manipulation operations to balance the workload in GPU. In addition, we used enhanced atomic operations in the latest NVIDIA CUDA architecture, Pascal, to sum up the pivot score, making the proposed algorithm can get efficient results when executing in parallel using the GPU.
|
author2 |
Wei-Mei Chen |
author_facet |
Wei-Mei Chen Tsung-Han Wu 吳宗翰 |
author |
Tsung-Han Wu 吳宗翰 |
spellingShingle |
Tsung-Han Wu 吳宗翰 A Study on the Problem of Maximal Clique Enumeration on GPU |
author_sort |
Tsung-Han Wu |
title |
A Study on the Problem of Maximal Clique Enumeration on GPU |
title_short |
A Study on the Problem of Maximal Clique Enumeration on GPU |
title_full |
A Study on the Problem of Maximal Clique Enumeration on GPU |
title_fullStr |
A Study on the Problem of Maximal Clique Enumeration on GPU |
title_full_unstemmed |
A Study on the Problem of Maximal Clique Enumeration on GPU |
title_sort |
study on the problem of maximal clique enumeration on gpu |
publishDate |
2017 |
url |
http://ndltd.ncl.edu.tw/handle/9374y3 |
work_keys_str_mv |
AT tsunghanwu astudyontheproblemofmaximalcliqueenumerationongpu AT wúzōnghàn astudyontheproblemofmaximalcliqueenumerationongpu AT tsunghanwu lièjǔjídàwánquánzitúwèntíyúgpuyùnsuànjiàgòuzhīyánjiū AT wúzōnghàn lièjǔjídàwánquánzitúwèntíyúgpuyùnsuànjiàgòuzhīyánjiū AT tsunghanwu studyontheproblemofmaximalcliqueenumerationongpu AT wúzōnghàn studyontheproblemofmaximalcliqueenumerationongpu |
_version_ |
1719153266138284032 |