readme

C++蚂蚁实现/C++蚂蚁实现 C++蚂蚁实现

README file for ant-based clustering as described in the paper 
Handl, Knowles and Dorigo (2004) Ant-based clustering and topographic 
mapping. Technical report TR-IRIDIA-2004-09.

To install (under Linux) follow the following steps:

1) Enter the directory ants/source and type "make accl".

2) To run the program for one of the predefined (hard-coded data sets)
go to the directory ants/bin and type "./accl <problemname>", where
<problemname> is e.g. square1 or sizes1. For a complete list of possible
problem names see the file ants/source/testset.C.

3) Output on the command line are the cluster sizes, the number of clusters,
different measures of clustering quality (F-Measure, Rand Index, Variance,
Dunn Index) and the runtime. Additional output files are the files
"initialdata.dat" (containing the generated data set) and 0.solution
(containing the clustering solution in the format <item no><cluster no>).

4) The algorithm can be repeatedly run on the same data set by
changing the parameter "evalnbr" in ants/include/conf.h. The 
accl_*.dat then store the algorithm's performance for each
run, and the accl_summary.dat file provides summary statistics
Mean and standard deviation).

5) The ant algorithm is implemented in the file accl.C.
Parameters can be changed in conf.h.

For feedback, questions and problems, please contact [e-mail].