rnsc
所属分类:Windows编程
开发工具:Unix_Linux
文件大小:101KB
下载次数:12
上传日期:2009-06-13 20:44:14
上 传 者:
ilhanmeng
说明: Restricted Neighborhood Search Cluster Algorithm, data mining ,聚类分析
(Restricted Neighborhood Search Cluster Algorithm, data mining, clustering analysis)
文件列表:
rnsc\definitions.h (117, 2009-06-11)
rnsc\linkedList3.cpp (6444, 2009-06-11)
rnsc\experiment.cpp~ (8631, 2004-10-18)
rnsc\graph2.cpp (12603, 2009-06-11)
rnsc\graph3.cpp~ (13000, 2004-10-18)
rnsc\graph5.cpp (15636, 2009-06-11)
rnsc\rnsc (64405, 2009-06-13)
rnsc\graph.cpp (15211, 2009-06-11)
rnsc\linkedList2.cpp (3974, 2009-06-11)
rnsc\out.rnsc (62, 2009-06-13)
rnsc\linkedList.h~ (2525, 2009-06-11)
rnsc\build (232, 2009-06-13)
rnsc\statsBook.h (430, 2009-06-11)
rnsc\graph3.cpp (11761, 2009-06-11)
rnsc\graph2.cpp~ (13842, 2004-10-18)
rnsc\graph4.cpp (14593, 2009-06-11)
rnsc\graph4.cpp~ (15833, 2004-10-18)
rnsc\printing.cpp~ (3919, 2004-10-18)
rnsc\linkedList2.cpp~ (5212, 2004-10-18)
rnsc\main.cpp~ (6750, 2004-10-18)
rnsc\graph (105, 2004-10-18)
rnsc\linkedList.cpp (4130, 2009-06-11)
rnsc\linkedList4.c (4111, 2009-06-11)
rnsc\linkedList4.c~ (5349, 2009-03-18)
rnsc\graph.h~ (6804, 2009-03-18)
rnsc\graph.cpp~ (16450, 2004-10-18)
rnsc\miscFunctions.cpp (2327, 2009-06-11)
rnsc\miscFunctions.h (243, 2009-06-11)
rnsc\printing.cpp (2681, 2009-06-11)
rnsc\experiment.cpp (7392, 2009-06-11)
rnsc\linkedList.cpp~ (5369, 2004-10-18)
rnsc\miscFunctions.cpp~ (3565, 2004-10-18)
rnsc\experiment.h (1321, 2009-06-11)
rnsc\statsBook.h~ (1668, 2009-03-18)
rnsc\linkedList3.cpp~ (7682, 2004-10-18)
rnsc\graph6.cpp~ (5641, 2004-10-18)
rnsc\graph6.cpp (4402, 2009-06-11)
rnsc\y2k.command (175, 2004-10-18)
... ...
NAME
rnsc - Restricted Neighbourhood Search Cluster Algorithm
SYNOPSIS
rnsc -g graph_filename
[-s] [-cnum] [-tnum] [-Tnum] [-nnum]
[-Nnum] [-enum] [-Dnum] [-dnum]
[-i input_clustering_filename]
[-o output_clustering_filename]
DESCRIPTION
The RNSC algorithm takes a simple graph as input and clusters it,
writing the final clustering to output_clustering_filename. There are
several user-input parameters that affect the algorithm's behaviour.
OPTIONS
-g graph_filename
RNSC reads the input graph from a filename, given as graph_filename.
The default filename is "graph". This file must have a very strict
format. The format is an adjacency list in which each edge appears
only once. The vertices are labelled with the integers 0, 1, ..., n-1.
The list of neighbours for vertex v appears as
v n_1 n_2 ... n_x -1
As an example, an instance of C_5, the cycle on four vertices, would
appear as
0 1 4 -1
1 2 -1
2 3 -1
3 4 -1
4 -1
-s
Skip the naive scheme and proceed directly to the scaled scheme in each
experiment. This option generally causes the algorithm to take longer,
but it is recommended when using an input clustering from a file.
-c num
Allow no more than "num" clusters. "num" must be between 2 and n,
where n is the number of vertices in the graph. If the -c option is
not specified or an invalid value is given, n clusters are used.
-t num
Set the tabu length to "num". Default value is 1. Note that when the
-T option is used, vertices can appear on the tabu list more than once
and moving them is only forbidden when they are on the tabu list more
than TabuTol times, where TabuTol is the tabu list tolerance.
-T num
Set the tabu list tolerance to "num". Default value is 1. The tabu
list tolerance is the number of times a vertex must appear on the tabu
list before moving it is forbidden.
-n num
Set the naive stopping tolerance to "num". Default value is 5. This
is the number of steps that the naive scheme will continue without
improving the best cost. If you run the scaled scheme, using a higher
naive stopping tolerance isn't likely to improve your results.
-N num
Set the scaled stopping tolerance to "num". Default value is 5. This
is the number of steps that the scaled scheme will continue without
improving the best cost. Setting the tolerance to 0 will cause the
algorithm to skip the scaled scheme.
-e num
Run "num" experiments. The best final clustering over all experiments
will be written to file. Default is 1.
-D num
Set the diversification frequency to "num". Without this option, no
diversification will be performed. If the -d flag is also used, then
"num" is the shuffling diversification frequency. If the -d flag is
not used, then "num" is the destructive diversification frequency. It
is recommended that the -d flag is used, because in the fast version of
RNSC (i.e. this one), destructive diversification isn't much help.
-d num
Set the shuffling diversification length to "num". That means that the
last "num" moves in the diversification period will be diversification
moves. Don't set this to be higher than the diversification frequency.
-i input_clustering_filename
Read an initial clustering from file. This file must have the same
format as an RNSC output clustering file. That is, each cluster is a
list of vertices, terminated with a -1 token. By default, the initial
clustering is random.
-o output_clustering_filename
Write the final clustering to a given filename. Default is out.rnsc.
EXAMPLE
rnsc -g ../graphs/11.gra -e3 -iin.rnsc -oout.rnsc -n2 -N100 -D40 -d10
-c300 -t15 -T2
SEE ALSO
Andrew D. King and Rudi Mathon. A fast cost-based graph clustering
algorithm. Manuscript.
Andrew D. King. Graph Clustering with Restricted Neighbourhood Search.
M. Sc. thesis, University of Toronto. 2004.
Andrew D. King, Natasa Przulj, and Igor Jurisica. Protein complex
prediction via cost-based clustering. Bioinformatics Advance Access.
2004.
近期下载者:
相关文件:
收藏者: