Minimal spanning Trees: Kruskal & Prim

Question

I am revising for a test and came across two questions relating to minimal spanning trees in graphs that I am unsure of and want to test my answers against.

The first one asks: if a graph has multiple minimal spanning trees, will Kruskal and Prim’s minimal spanning tree algorithms generate the same tree? I am thinking that they won’t necessarily because the algorithms are different. Kruskal relies on the edges being sorted by weight, whilst Prim doesn’t, so they could start at different vertices, thus generating different trees.

The second question asks: if a graph has multiple minimal spanning trees, how should Kruskal’s algorithm be adapted to generate all of these? I am thinking that one needs to allow for either a looping structure through the vertices so that the start vertex changes each time because the edges values may be the same. Thus generate trees by taking each of the vertices in turn as start vertex. In other words, sort by vertex numbering as well, not by edge weight alone.

score 1 · Accepted Answer · edited May 23 '17 at 11:45

If a graph has multiple minimal spanning trees, will Kruskal and Prim’s minimal spanning tree algorithms generate the same tree?

No, it is not necessary for the Prims and Kruskals algorithms to generate the same MST. A graph can have many MSTs and both algorithms can generate different ones. But the edge types of the two MSTs will necessarily be the same. ie, if you make a multiset of the edges of the two MSTs, then the two multisets will definitely be equal. You can find a proof of this here

If a graph has multiple minimal spanning trees, how should Kruskal’s algorithm be adapted to generate all of these?

There seems to be no direct reduction of the kruskal's MST algorithm to find all the MSTs in the graph. Your best bet will be

Step1 : Sort the edges of the graph as done in kruskals.

Step2 : Now for each edge in the sorted list, two things can happen. Either the edge is in an MST or it is not. So for each edge in the sorted list, we will go over these two cases and create two new Union-Find data structures and recurse on other edges.

pseudocode:

Step1: sort edges in ascending order
Step2: now call printAllMsts(0, new UnionFind(V))

void printAllMsts(int edgeNum, UnionFind U){
    if(edgeNum == edges.length){  // If no more edges to add
        if(U.numEdges == V-1){    // If U has V-1 edges, then we have an MST
             printMst();
        }
        return;
    }

    if(edges[edgeNum+1] == edges[edgeNum]){
         printAllMsts(edgeNum+1, U); // when E is not taken in the MST   
    }

    edge E = edges[edgeNum];
    If(E can be a part of some MST){
        UnionFind newU = new UnionFind(U);
        newU.add(E);
    }

    printAllMsts(edgeNum+1, newU);
}

The running time of the algorithm will depend on the number and type of edges in the graph. The worst case input for the problem will be when all the edges in the graph are of the same length. The running time is atleast O(V*numberOfMsts) because whenever there is a possibility of different MSTs, the current Union-Find data structure is cloned which takes O(V) time.

Minimal spanning Trees: Kruskal & Prim

1 Answers1