1 Minimum Spanning Trees (MSTs) and Minimum Cost Spanning Trees (MCSTs) What is a Minimum Spanning Tree? Constructing Minimum Spanning Trees. What is a Minimum-Cost Spanning Tree (MCST) ? Applications of Minimum Cost Spanning Trees. MCST Algorithms: –Kruskal’s Algorithm. Tracing Kruskal’s Algorithm Implementation. –Prim’s algorithm. Tracing Prim’s Algorithm Implementation. Review Questions.

2 What is a Minimum Spanning Tree? Let G = (V, E) be a simple, connected, undirected graph that is not edge-weighted. A spanning tree of G is a free tree (i.e., a tree with no root) with | V | - 1 edges that connects all the vertices of the graph. Thus a minimum spanning tree T = (V’, E’) for G is a subgraph of G with the following properties:  V’ = V  T is connected  T is acyclic (  |E’| = |V| - 1) A spanning tree is called a tree because every acyclic undirected graph can be viewed as a general, unordered tree. Because the edges are undirected, any vertex may be chosen to serve as the root of the tree.

3 Constructing Minimum Spanning Trees Any traversal of a connected, undirected graph visits all the vertices in that graph. The set of edges which are traversed during a traversal forms a spanning tree. For example, Fig (b) shows a spanning tree of G obtained from a breadth-first traversal starting at vertex a. Similarly, Fig (c) shows a spanning tree of G obtained from a depth-first traversal starting at vertex c. (a) Graph G (b) Breadth-first spanning tree of G starting from a (c) Depth-first spanning tree of G starting from c

4 What is a Minimum-Cost Spanning Tree? For an edge-weighted, connected, undirected graph, G, the total cost of G is the sum of the weights on all its edges. A minimum-cost spanning tree for G is a minimum spanning tree of G that has the least total cost. Example: The graph Has 16 spanning trees. Some are: The graph has two minimum-cost spanning trees, each with a cost of 6:

5 Applications of Minimum-Cost Spanning Trees Minimum-cost spanning trees have many applications. Some are: Building cable networks that join n locations with minimum cost. Building a road network that joins n cities with minimum cost. Obtaining an independent set of circuit equations for an electrical network. In pattern recognition minimal spanning trees can be used to find noisy pixels.

6 MCST Algorithms: Kruskal's Algorithm Kruskal’s algorithm finds the minimum cost spanning tree of a graph by adding edges one-by-one to an initially empty free tree enqueue edges of G in a queue in increasing order of cost. T =  ; while(queue is not empty){ dequeue an edge e; if(e does not create a cycle with edges in T) add e to T; } return T; Running time is O(E log V)

7 Tracing Kruskal’s Algorithm Trace Kruskal's algorithm in finding a minimum-cost spanning tree for the undirected, weighted graph given below: The minimum cost is: 24

8 Implementation of Kruskal's Algorithm public static Graph kruskalsAlgorithm(Graph g){ Graph result = new GraphAsLists(false); Iterator it = g.getVertices(); while (it.hasNext()){ Vertex v = (Vertex)it.next(); result.addVertex(v.getLabel()); } PriorityQueue queue = new BinaryHeap(g.getNumberOfEdges()); it = g.getEdges(); while(it.hasNext()){ Edge e = (Edge) it.next(); if (e.getWeight()==null) throw new IllegalArgumentException("Graph is not weighted"); queue.enqueue(e); } while (!queue.isEmpty()){ // test if there is a path from vertex // from to vertex to before adding the edge Edge e = (Edge) queue.dequeueMin(); String from = e.getFromVertex().getLabel(); String to = e.getToVertex().getLabel(); if (!result.isReachable(from, to)) result.addEdge(from,to,e.getWeight()); } return result; } adds an edge only, if it does not create a cycle

9 Implementation of Kruskal's Algorithm (Cont’d) public abstract class AbstractGraph implements Graph { public boolean isReachable(String from, String to){ Vertex fromVertex = getVertex(from); Vertex toVertex = getVertex(to); if (fromVertex == null || toVertex==null) throw new IllegalArgumentException("Vertex not in the graph"); PathVisitor visitor = new PathVisitor(toVertex); this.preorderDepthFirstTraversal(visitor, fromVertex); return visitor.isReached(); } private class PathVisitor implements Visitor { boolean reached = false; Vertex target; PathVisitor(Vertex t){target = t;} public void visit(Object obj){ Vertex v = (Vertex) obj; if (v.equals(target)) reached = true; } public boolean isDone(){return reached;} boolean isReached(){return reached;} }

10 MCST Algorithms: Prim’s Algorithm The algorithm continuously increases the size of a tree starting with a single vertex until it spans all the vertices of G: Consider a weighted, connected, undirected graph G=(V, E) and a free tree T = (V T, E T ) that is a subgraph of G Initialize: V T = {x}; // where x is an arbitrary vertex from V Initialize: E T =  ; repeat{ Choose an edge (v,w) with minimal weight such that v is in V T and w is not (if there are multiple edges with the same weight, choose arbitrarily but consistently); // add an minimal weight edge that will not form a cycle Add vertex v to V T ; add edge (v, w) to E T; }until (V T == V); Time complexity (total)Graph and Minimum edge weight data structure O(V 2 )adjacency matrix and array O((V + E) log(V)) = O(E log(V))adjacency list and binary heap O(E + V log(V))Adjacency list and Fibonacci heap Time Complexity of Prim’s Algorithm:

11 Tracing Prim’s Algorithm (Example adopted from Wikipedia) We want to find a Minimum Cost Spanning Tree for the graph on the left. The numbers near the edges indicate the edge weights. Vertex D has been arbitrarily chosen as a starting point. Vertices A, B, E and F are connected to D through a single edge. A is the vertex nearest to D and will be chosen as the second vertex along with the edge AD. The next vertex chosen is the vertex nearest to either D or A. B is 9 away from D and 7 away from A, E is 15, and F is 6. F is the smallest distance away, so we highlight the vertex F and the arc DF.

12 Tracing Prim’s Algorithm (Cont’d) The algorithm carries on as above. Vertex B, which is 7 away from A, is highlighted. In this case, we can choose between C, E, and G. C is 8 away from B, E is 7 away from B, and G is 11 away from F. E is nearest, so we highlight the vertex E and the arc BE. Here, the only vertices available are C and G. C is 5 away from E, and G is 9 away from E. C is chosen, so it is highlighted along with the arc EC.

13 Tracing Prim’s Algorithm (Cont’d) Vertex G is the only remaining vertex. It is 11 away from F, and 9 away from E. E is nearer, so we highlight it and the arc EG. Now all the vertices have been selected and the minimum cost spanning tree is shown in green. In this case, it has weight 39.

14 Tracing Prim’s Algorithm: An alternative method Trace Prim’s algorithm starting at vertex a: The resulting minimum-cost spanning tree is: Note: Current Active vertex is vertex in the row with minimum weight from the previous column, v1 is the active vertex in the column where weight appears for the first time

15 Implementation of Prim’s Algorithm using MinHeap The implementation of Prim's algorithm uses an Entry class, which contains the following three fields: –know: a boolean variable indicating whether the weight of the edge from this vertex to an active vertex v1 is known, initially false for all vertices. –weight : the minimum known weight from this vertex to an active vertex v1, initially infinity for all vertices except that of the starting vertex which is 0. –predecessor : the predecessor of an active vertex v1 in the MCST, initially unknown for all vertices. public class Algorithms{ static final class Entry{ boolean known; int weight; Vertex predecessor; Entry(){ known = false; weight = Integer.MAX_VALUE; predecessor = null; }

16 Implementation of Prim’s Algorithm using MinHeap (Cont’d) make a table of values (known, weight(edge), predecessor) initially (FALSE, ,null) for each vertex except the start vertex table entry is set to (FALSE, 0, null) ; MinHeap =  ; MinHeap.enqueue( Association (0, startVertex)); // let tree T be empty; while (MinHeap is not empty) { // T has fewer than n vertices dequeue an Association (weight(e), v) from the MinHeap; Update table entry for v to (TRUE, weight(e), v); // add v and e to T for( each emanating edge f = (v, u) of v){ if( table[u].known == FALSE ) // u is not already in T if (weight(f) < table[u].weight) { // find the current min edge weight table[u].weight = weight(f); table[u].predecessor = v; enqueue(Association(weight(f), u); } The idea is to use a table array to remember, for each vertex, the smallest edge connecting T with that vertex. Each entry in table represents a vertex x. The object stored at table entry index(x) will have the predecessor of x and the corresponding edge weight.

17 Implementation of Prim’s Algorithm using MinHeap (Cont’d) Prim’s algorithm can be implemented as shown below: public static Graph primsAlgorithm(Graph g, Vertex start){ int n = g.getNumberOfVertices(); Entry table[] = new Entry[n]; for(int v = 0; v < n; v++) table[v] = new Entry(); table[g.getIndex(start)].weight = 0; PriorityQueue queue = new BinaryHeap(g.getNumberOfEdges()); queue.enqueue(new Association(new Integer(0), start)); while(!queue.isEmpty()) { Association association = (Association)queue.dequeueMin(); Vertex v1 = (Vertex) association.getValue(); int n1 = g.getIndex(v1); if(!table[n1].known){ table[n1].known = true; Iterator p = v1.getEmanatingEdges(); while (p.hasNext()){ Edge edge = (Edge) p.next(); Vertex v2 = edge.getMate(v1); int n2 = g.getIndex(v2); Integer weight = (Integer) edge.getWeight(); int d = weight.intValue();

18 Implementation of Prim’s Algorithm using MinHeap (Cont'd) if(!table[n2].known && d < table[n2].weight){ table[n2].weight = d; table[n2].predecessor = v1; queue.enqueue(new Association(new Integer(d), v2)); } GraphAsLists result = new GraphAsLists(false); Iterator it = g.getVertices(); //add vertices of this graph to result while (it.hasNext()){ Vertex v = (Vertex) it.next(); result.addVertex(v.getLabel()); } // add edges to result using each vertex predecessor in table it = g.getVertices(); while (it.hasNext()){// Vertex v = (Vertex) it.next(); if (v != start){ int index = g.getIndex(v); String from = v.getLabel(); String to = table[index].predecessor.getLabel(); result.addEdge(from, to, new Integer(table[index].distance)); } return result; }

19 Tracing Prim’s Algorithm from the MinHeap implementation Trace Prim’s algorithm in finding a MCST starting from vertex A: Enqueue edge weights for edges emanating from A Enqueue edge weights for edges emanating from C

20 Tracing Prim’s Algorithm from the MinHeap implementation (Cont’d) The MCST:

21 Note on MCSTs generated by Prim’s and Kruskal’s Algorithms Note: It is not necessary that Prim's and Kruskal's algorithm generate the same minimum-cost spanning tree. For example for the graph: Kruskal's algorithm (that imposes an ordering on edges with equal weights) results in the following minimum cost spanning tree: The same tree is generated by Prim's algorithm if the start vertex is any of: A, B, or D; however if the start vertex is C the minimum cost spanning tree is:

22 Review Questions 1.Find the breadth-first spanning tree and depth-first spanning tree of the graph G A shown above. 2.For the graph G B shown above, trace the execution of Prim's algorithm as it finds the minimum-cost spanning tree of the graph starting from vertex a. 3.Repeat question 2 above using Kruskal's algorithm. GBGB