Connected graph given adjacency matrix
11 views (last 30 days)
Show older comments
imperial1991
on 29 May 2012
Answered: Hon Wah Yeung
on 31 Jan 2021
Hi all, I'm working on a research project on graphical models involving a large dimension (large number of nodes). I'm just wondering, is there an existing efficient algorithm to determine whether the graph is connected or not given its adjacency matrix? I've tried looking but it seems that the existing ones are brute force algorithms which are costly. Does MATLAB have a built-in function? When I checked, it seems that none exists as of now.
Help would be greatly appreciated!
0 Comments
Accepted Answer
Wolfgang Schwanghart
on 29 May 2012
You can use the function dmperm to see if a graph consists of one or several connected components. E.g. see the example here http://blogs.mathworks.com/steve/2007/03/20/connected-component-labeling-part-3/
HTH, W.
More Answers (3)
Christine Tobler
on 22 Dec 2016
Edited: Christine Tobler
on 22 Nov 2019
I realize this is an old question, but since it's still getting visits, I have a small addition. As of R2015b, the new graph and digraph classes have a method for computing connected components. To check whether a graph is connected based on its adjacency matrix A, use
g = digraph(A);
bins = conncomp(g, 'Type', 'weak');
isConnected = all(bins == 1);
The vector bins gives the bin number for each node of A. If a graph is connected, all nodes will be in one bin, which is checked using all(bins == 1). This is not necessarily faster than dmperm, but easier to read.
2 Comments
Jonathan DuBois
on 24 Aug 2017
What is the point of the first command? Should it be: bins = conncomp(g, 'Type', 'weak')? It works with g in the second command but with A, I get the following error: Undefined function 'conncomp' for input arguments of type 'double'.
Christine Tobler
on 22 Nov 2019
Thank you, yes, this was a typo. I've edited the original answer to fix this (although it took me two years to realize you had commented on this answer).
Hon Wah Yeung
on 31 Jan 2021
This will work if you can at least load the matrix (meaning the matrix is not larger than the max possible array for Matlab) to Matlab.
function [idx,Group] = CheckConnected(E)
L=length(E);
T=find(E(:,1)~=0);
T=[T;1]';
M=sum(E(:,T),2);
T1=union(find(M~=0),T);
while length(T1)>length(T)
T=T1;
M=sum(E(:,T),2);
T1=union(find(M~=0),T);
end
if length(T)==L
idx=1;
Group=T;
else
idx=0;
Group=T;
end
end
0 Comments
See Also
Categories
Find more on 2-D and 3-D Plots in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!