ECE 250 Algorithms and Data Structures

Prim’s algorithm

Douglas Wilhelm Harder, M.Math. LEL

Department of Electrical and Computer Engineering
University of Waterloo
Waterloo, Ontario, Canada

ece.uwaterloo.ca
dwharder@alumni.uwaterloo.ca

© 2006-2013 by Douglas Wilhelm Harder. Some rights reserved.

 Prim’s algorithm
2

Outline

This topic covers Prim’s algorithm:

– Finding a minimum spanning tree
– The idea and the algorithm
– An example
 Prim’s algorithm
3

Strategy

Suppose we take a vertex

– Given a single vertex e1, it forms a minimum spanning tree on one
vertex

v1
 Prim’s algorithm
4

Strategy

Add that adjacent vertex v2 that has a connecting edge e1 of

minimum weight
– This forms a minimum spanning tree on our two vertices and e1 must be
in any minimum spanning tree containing the vertices v1 and v2

v2

e1

v1
 Prim’s algorithm
5

Strategy

Strategy:
– Suppose we have a known minimum spanning tree on k < n vertices
– How could we extend this minimum spanning tree?
 Prim’s algorithm
6

Strategy

Add that edge ek with least weight that connects this minimum
spanning tree to a new vertex vk + 1
– This does create a minimum spanning tree on k + 1 nodes—there is no
other edge we could add that would connect this vertex
– Does the new edge, however, belong to the minimum spanning tree on
all n vertices?

vk + 1
ek
 Prim’s algorithm
7

Strategy

Suppose it does not

– Thus, vertex vk + 1 is connected to the minimum spanning tree via another
sequence of edges

vk + 1
ek
 Prim’s algorithm
8

Strategy

Because a minimum spanning tree is connected, there must be a

path from vertex vk + 1 back to our existing minimum spanning tree
– It must be connected along some edge e

vk + 1
ek

e
 Prim’s algorithm
9

Strategy

Let w be the weight of this minimum spanning tree

– Recall, however, that when we chose to add vk + 1, it was because ek was
the edge connecting an adjacent vertex with least weight
– Therefore e  ek where |e| represents the weight of the edge e

ek  e  0

vk + 1
ek

e
 Prim’s algorithm
10

Strategy

Consider, however, suppose we swap edges and instead choose to

include ek and exclude e
– The result is still a minimum spanning tree, but the weight is now
w  ek 1  e w

vk + 1
ek + 1

e
 Prim’s algorithm
11

Strategy

Thus, by swapping ek for e , we have a spanning tree that has less

weight than the so-called minimum spanning tree containing e
– This contradicts our assumption that the spanning tree containing e
was minimal
– Therefore, our minimum spanning tree must contain ek

vk + 1
ek

e
 Prim’s algorithm
12

Strategy

Recall that we did not prescribe the value of k, and thus, k could be
any value, including k = 1

vk + 1
ek

e
 Prim’s algorithm
13

Strategy

Recall that we did not prescribe the value of k, and thus, k could be
any value, including k = 1
– Given a single vertex e1, it forms a minimum spanning tree on one
vertex

v1
 Prim’s algorithm
14

Strategy

Add that adjacent vertex v2 that has a connecting edge e1 of

minimum weight
– This forms a minimum spanning tree on our two vertices and e1 must be
in any minimum spanning tree containing the vertices v1 and v2

v2

e1

v1
 Prim’s algorithm
15

Minimum Spanning Trees

Prim’s algorithm for finding the minimum spanning tree states:

– Start with an arbitrary vertex to form a minimum spanning tree on one
vertex
– At each step, add that vertex v not yet in the minimum spanning tree
that has an edge with least weight that connects v to the existing
minimum spanning sub-tree
– Continue until we have n – 1 edges and n vertices

Another possibility is Kruskal’s algorithm

 Prim’s algorithm
16

Prim’s Algorithm

Associate with each vertex three items of data:

– A Boolean flag indicating if the vertex has been visited,
– The minimum distance to the partially constructed tree, and
– A pointer to that vertex which will form the parent node in the resulting
tree

For example:
– Add three member variables to the vertex class
– Track three tables
 Prim’s algorithm
17

Prim’s Algorithm

Initialization:
– Select a root node and set its distance as 0
– Set the distance to all other vertices as ∞
– Set all vertices to being unvisited
– Set the parent pointer of all vertices to 0
 Prim’s algorithm
18

Prim’s Algorithm

Iterate while there exists an unvisited vertex with distance < ∞

– Select that unvisited vertex with minimum distance
– Mark that vertex as having been visited
– For each adjacent vertex, if the weight of the connecting edge is less
than the current distance to that vertex:
• Update the distance to equal the weight of the edge
• Set the current vertex as the parent of the adjacent vertex
 Prim’s algorithm
19

Prim’s Algorithm

Halting Conditions:
– There are no unvisited vertices which have a distance < ∞

If all vertices have been visited, we have a spanning tree of the

entire graph

If there are vertices with distance ∞, then the graph is not connected
and we only have a minimum spanning tree of the connected sub-
graph containing the root
 Prim’s algorithm
20

Prim’s Algorithm

Let us find the minimum spanning tree for the following undirected
weighted graph
 Prim’s algorithm
21

Prim’s Algorithm

First we set up the appropriate table and initialize it

Distance Parent
1 F 0 0
2 F ∞ 0
3 F ∞ 0
4 F ∞ 0
5 F ∞ 0
6 F ∞ 0
7 F ∞ 0
8 F ∞ 0
9 F ∞ 0
 Prim’s algorithm
22

Prim’s Algorithm

Visiting vertex 1, we update vertices 2, 4, and 5

Distance Parent
1 T 0 0
2 F 4 1
3 F ∞ 0
4 F 1 1
5 F 8 1
6 F ∞ 0
7 F ∞ 0
8 F ∞ 0
9 F ∞ 0
 Prim’s algorithm
23

Prim’s Algorithm

What these numbers really mean is that at this point, we could

extend the trivial tree containing just the root node by one of the
three possible children:

As we wish to find a minimum spanning tree, it makes sense we add

that vertex with a connecting edge with least weight
 Prim’s algorithm
24

Prim’s Algorithm

The next unvisited vertex with minimum distance is vertex 4

– Update vertices 2, 7, 8
– Don’t update vertex 5

Distance Parent
1 T 0 0
2 F 2 4
3 F ∞ 0
4 T 1 1
5 F 8 1
6 F ∞ 0
7 F 9 4
8 F 8 4
9 F ∞ 0
 Prim’s algorithm
25

Prim’s Algorithm

Now that we have updated all vertices adjacent to vertex 4, we can

extend the tree by adding one of the edges
(1, 5), (4, 2), (4, 7), or (4, 8)

We add that edge with the least

weight: (4, 2)
 Prim’s algorithm
26

Prim’s Algorithm

Next visit vertex 2

– Update 3, 5, and 6

Distance Parent
1 T 0 0
2 T 2 4
3 F 2 2
4 T 1 1
5 F 6 2
6 F 1 2
7 F 9 4
8 F 8 4
9 F ∞ 0
 Prim’s algorithm
27

Prim’s Algorithm

Again looking at the shortest edges to each of the vertices adjacent

to the current tree, we note that we can add (2, 6) with the least
increase in weight
 Prim’s algorithm
28

Prim’s Algorithm

Next, we visit vertex 6:

– update vertices 5, 8, and 9

Distance Parent
1 T 0 0
2 T 2 4
3 F 2 2
4 T 1 1
5 F 3 6
6 T 1 2
7 F 9 4
8 F 7 6
9 F 8 6
 Prim’s algorithm
29

Prim’s Algorithm

The edge with least weight is (2, 3)

– This adds the weight of 2 to the weight minimum spanning tree
 Prim’s algorithm
30

Prim’s Algorithm

Next, we visit vertex 3 and update 5

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 F 2 3
6 T 1 2
7 F 9 4
8 F 7 6
9 F 8 6
 Prim’s algorithm
31

Prim’s Algorithm

At this point, we can extend the tree by adding the edge (3, 5)
 Prim’s algorithm
32

Prim’s Algorithm

Visiting vertex 5, we update 7, 8, 9

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 T 2 3
6 T 1 2
7 F 4 5
8 F 1 5
9 F 5 5
 Prim’s algorithm
33

Prim’s Algorithm

At this point, there are three possible edges which we could include
which will extend the tree

The edge to 8 has the least weight

 Prim’s algorithm
34

Prim’s Algorithm

Visiting vertex 8, we only update vertex 9

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 T 2 3
6 T 1 2
7 F 4 5
8 T 1 5
9 F 3 8
 Prim’s algorithm
35

Prim’s Algorithm

There are no other vertices to update while visiting vertex 9

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 T 2 3
6 T 1 2
7 F 4 5
8 T 1 5
9 T 3 8
 Prim’s algorithm
36

Prim’s Algorithm

And neither are there any vertices to update when visiting vertex 7

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 T 2 3
6 T 1 2
7 T 4 5
8 T 1 5
9 T 3 8
 Prim’s algorithm
37

Prim’s Algorithm

At this point, there are no more unvisited vertices, and therefore we

are done

If at any point, all remaining vertices had a distance of ∞, this would

indicate that the graph is not connected
– in this case, the minimum spanning tree would only span one connected
sub-graph
 Prim’s algorithm
38

Prim’s Algorithm

Using the parent pointers, we can now construct the minimum

spanning tree

Distance Parent
1 T 0 0
2 T 2 4
3 T 2 2
4 T 1 1
5 T 2 3
6 T 1 2
7 T 4 5
8 T 1 5
9 T 3 8
 Prim’s algorithm
39

Prim’s Algorithm

To summarize:
– we begin with a vertex which represents the root
– starting with this trivial tree and iteration, we find the shortest edge
which we can add to this already existing tree to expand it

This is a reasonably efficient algorithm: the number of visits to

vertices is kept to a minimum
 Prim’s algorithm
40

Implementation and analysis

The initialization requires Q(|V|) memory and run time

We iterate |V| – 1 times, each time finding the closest vertex

– Iterating through the table requires is Q(|V|) time
– Each time we find a vertex, we must check all of its neighbors
– With an adjacency matrix, the run time is Q(|V|(|V| + |V|)) = Q(|V|2)
– With an adjacency list, the run time is Q(|V|2 + |E|) = Q(|V|2) as |E| = O(|V|2)

Can we do better?
– Recall, we only need the shortest edge next
– How about a priority queue?
• Assume we are using a binary heap
• We will have to update the heap structure—this requires additional work
 Prim’s algorithm
41

Implementation and analysis

The initialization still requires Q(|V|) memory and run time

– The priority queue will also requires O(|V|) memory

We iterate |V| – 1 times, each time finding the closest vertex

– Place the shortest distances into a priority queue
– The size of the priority queue is O(|V|)
– Thus, the work required for this is O(|V| ln(|V|))

Is this all the work that is necessary?

– Recall that each edge visited may result in a new edge being pushed to
the very top of the heap
– Thus, the work required for this is O(|E| ln(|V|))

Thus, the total run time is O(|V| ln(|V|) + |E| ln(|V|)) = O(|E| ln(|V|))
 Prim’s algorithm
42

Implementation and analysis

Here is a worst-case graph if we were to start with Vertex A

– Assume that the adjacency lists are in order
– Each time, the edge is percolated to the top of the heap
 Prim’s algorithm
43

Implementation and analysis

We could use a different heap structure:

– A Fibonacci heap is a node-based heap
– Pop is still O(ln(|V|)), but inserting and moving a key is Q(1)
– Thus, because we are only calling pop |V| – 1 times, work required
reduces to O(|E| + |V| ln(|V|))
– Thus, the overall run-time is O(|E| + |V| ln(|V|))
 Prim’s algorithm
44

Implementation and analysis

Thus, we have two run times when using

– A binary heap: O(|E| ln(|V|))
– A Fibonacci heap:O(|E| + |V| ln(|V|))

Questions: Which is faster if |E| = Q(|V|)? How about if |E| = Q(|V|2)?

 Prim’s algorithm
45

Summary

We have seen an algorithm for finding minimum spanning trees

– Start with a trivial minimum spanning tree and grow it
– An alternate algorithm, Kruskal’s algorithm, uses a different approach

Prim’s algorithm finds an edge with least weight which grows an

already existing tree
– It solves the problem in O(|E| ln(|V|)) time
 Prim’s algorithm
46

References

Wikipedia, http://en.wikipedia.org/wiki/Minimum_spanning_tree
Wikipedia, http://en.wikipedia.org/wiki/Prim’s_algorithm

These slides are provided for the ECE 250 Algorithms and Data Structures course. The
material in it reflects Douglas W. Harder’s best judgment in light of the information available to
him at the time of preparation. Any reliance on these course slides by any party for any other
purpose are the responsibility of such parties. Douglas W. Harder accepts no responsibility for
damages, if any, suffered by any party as a result of decisions made or actions based on these
course slides for any other purpose than that for which it was intended.