pagerank

A pagerank implementation in C++ able to handle very big graphs

  • 所有者: louridas/pagerank
  • 平台:
  • 许可证:
  • 分类:
  • 主题:
  • 喜欢:
    0
      比较:

Github星跟踪图

An Open Source PageRank Implementation

This project provides an open source PageRank implementation. The
implementation is a straightforward application of the algorithm
description given in the American Mathematical Society's Feature
Column How Google Finds Your Needle in the Web's
Haystack
,
by David Austing. It can handle very big hyperlink graphs with
millions of vertices and arcs.

Building

The project is written in standard C++ and can be built by running:

g++ -o pagerank pagerank.cpp table.cpp table.h

Usage

pagerank is invoked by

pagerank [OPTIONS] graph_file

where OPTIONS may be:

  • -t: if set, tracing is enabled. Tracing outputs all intermediate
    steps of the pagerank calculation algorithm and can be very
    voluminous.

  • -n: if set, the graph_file is in numeric format, that is it consists
    of lines of the form <from><delim><to> where <from> and <to> are
    integer vertex indices, starting with zero. If not set, the
    graph_file consists of lines of the form <from><delim><to> where
    <from> and <to> are vertex IDs that will be interpreted as strings.

  • -a <float>: the pagerank dumping factor; default is 0.85.

  • -c <float>: the convergence criterion. The pagerank iterations will
    stop when two successive iterations have converged to less than or
    equal to this value. Default is 0.00001.

  • -s <integer>: the number of rows of the hyperlink matrix. This is
    not the maximum size; if the graph requires more rows, they will be
    allocated as necessary. If an approximate size is known beforehand,
    however, the necessary internal data structures can be
    pre-allocated for better performance.

  • -d <string>: the delimited used to separate vector indices in the
    input graph file. Default is " => ".

Testing

Testing the implementation was carried out by comparing with pagerank
calculations produced by Jung. The
Java code that produced the calculations can be found in the java
directory of the project. The test suites can be found at the test
directory of the project, along with the test driver program
pagerank_test.cpp,
which is invoked by: pagerank_test <test_suite>.

The <test_suite> is a file containing in each line a filename, in the
same directory, with an input graph. For an input graph foo.txt, the
pagerank results against which we want to compare our implementation
is found in foo-pr.txt. The result files are generated by the
pagerank_calc.sh script.

The test driver is written in standard C++ and can be compiled with:

g++ -I../cpp -o pagerank_test pagerank_test.cpp ../cpp/table.cpp 

The graph test files were generated by the
igraph R port using the R scripts in
the test directory.

主要指标

概览
名称与所有者louridas/pagerank
主编程语言C++
编程语言R (语言数: 7)
平台
许可证
所有者活动
创建于2012-05-22 06:44:34
推送于2015-01-02 08:02:58
最后一次提交2015-01-02 09:02:58
发布数0
用户参与
星数279
关注者数12
派生数132
提交数14
已启用问题?
问题数9
打开的问题数4
拉请求数2
打开的拉请求数1
关闭的拉请求数0
项目设置
已启用Wiki?
已存档?
是复刻?
已锁定?
是镜像?
是私有?