Problem with pmetis
I’ve had a problem during the use of Pmetis and Kmetis; I executed Pmetis and Kmetis 4.0 on windows using as input the example graph in the figure No.8c of the manual.
The input of the example graph in the figure No.8c is the following:
7 11 11
4 5 1 3 2 2 1
2 1 1 3 2 4 1
5 5 3 4 2 2 2 1 2
3 2 1 3 2 6 2 7 5
1 1 1 3 3 6 2
6 5 2 4 2 7 6
2 6 6 4 5
Using the graph checker provided by Metis, I observed that the graph
isn’t valid.
For this reason I added a digit 1 in the first row so the input file
which I used is the following:
7 11 11 1
4 5 1 3 2 2 1
2 1 1 3 2 4 1
5 5 3 4 2 2 2 1 2
3 2 1 3 2 6 2 7 5
1 1 1 3 3 6 2
6 5 2 4 2 7 6
2 6 6 4 5
Using Pmetis to obtain five partitions of this graph (using the command line: pmetis graphFileName 5) , I obtained the
following result:
3
3
3
0
4
0
0
I think that the obtained result is formed only by three partitions instead of five. Did I do any mistake?
Submitted by rarefatto on Tue, 2009-07-28 13:19
»
- Login to post comments
RE: I read a previous post that
I read a previous post that maybe contains the answer to my post
"The algorithms in the current version of Metis are not optimized for "over-partitioning" small graphs. This is something that will be added in the 5.0 release, which is currently under development."
If this is the correct answer, do you know other algorithm/library that can be used on Windows that allow to do "over-partitioning" on small graphs(i work with graph with about 100 vertices)