comparison toolboxes/FullBNT-1.0.7/graph/minimum_spanning_tree.m @ 0:e9a9cd732c1e tip

first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 function A = minimum_spanning_tree(C1, C2)
2 %
3 % Find the minimum spanning tree using Prim's algorithm.
4 % C1(i,j) is the primary cost of connecting i to j.
5 % C2(i,j) is the (optional) secondary cost of connecting i to j, used to break ties.
6 % We assume that absent edges have 0 cost.
7 % To find the maximum spanning tree, used -1*C.
8 % See Aho, Hopcroft & Ullman 1983, "Data structures and algorithms", p 237.
9
10 % Prim's is O(V^2). Kruskal's algorithm is O(E log E) and hence is more efficient
11 % for sparse graphs, but is implemented in terms of a priority queue.
12
13 % We partition the nodes into those in U and those not in U.
14 % closest(i) is the vertex in U that is closest to i in V-U.
15 % lowcost(i) is the cost of the edge (i, closest(i)), or infinity is i has been used.
16 % In Aho, they say C(i,j) should be "some appropriate large value" if the edge is missing.
17 % We set it to infinity.
18 % However, since lowcost is initialized from C, we must distinguish absent edges from used nodes.
19
20 n = length(C1);
21 if nargin==1, C2 = zeros(n); end
22 A = zeros(n);
23
24 closest = ones(1,n);
25 used = zeros(1,n); % contains the members of U
26 used(1) = 1; % start with node 1
27 C1(find(C1==0))=inf;
28 C2(find(C2==0))=inf;
29 lowcost1 = C1(1,:);
30 lowcost2 = C2(1,:);
31
32 for i=2:n
33 ks = find(lowcost1==min(lowcost1));
34 k = ks(argmin(lowcost2(ks)));
35 A(k, closest(k)) = 1;
36 A(closest(k), k) = 1;
37 lowcost1(k) = inf;
38 lowcost2(k) = inf;
39 used(k) = 1;
40 NU = find(used==0);
41 for ji=1:length(NU)
42 for j=NU(ji)
43 if C1(k,j) < lowcost1(j)
44 lowcost1(j) = C1(k,j);
45 lowcost2(j) = C2(k,j);
46 closest(j) = k;
47 end
48 end
49 end
50 end
51