11/2/2018

Prim’s Algorithm Example

More Proofs
Application of MST to Clustering
CSCI 3100

Prim’s algorithm with priority queue

1
 11/2/2018

Prim’s algorithm example: r = 1

A A.key =0 A.π = NIL
6 5 B.key = ∞ B.π = NIL
1 C.key = ∞ C.π = NIL
C
B D.key = ∞ D.π = NIL
5 5
D E.key = ∞ E.π = NIL
3 6 2 F.key = ∞ F.π = NIL
4
Q=ABCDEF
E F u = EXTRACT_MIN(Q)
6
u=A

Prim’s algorithm example: r = 1

Q=BCDEF
For each v in G.Adg(A)
A if v in Q and w (u, v) < v.key
6 5 v. π = u
1 v.key = w(u,v)
C
B (A, B): B is in Q, w(A,B) < B.key
5 5
D B. π = A, B.key = 6, Q = B C D E F
3 6 2 (A, D): D is in Q, w(A,D) < D.key
4
D. π = A, D.key = 1, Q = D B C E F
E F (A, C): C is in Q, w(A, C) < C.key
6 C. π = A, C.key = 5, Q = D B C E F

2
 11/2/2018

B.key = 6, D.key = 1, C.key = 5, E.key = ∞,

F.key = ∞
Q=DBCEF
u = EXTRACT_MIN(Q) = D
A (D, A): A is not in Q
6 5 (D, B): B is in Q, w(D,B) < B.key
1 C B. π = D, B.key = 5, Q = B C E F
B
5 5 (D, C): C is in Q, w(D, C) == C.key
D (D, E): E is in Q, w(D, E) < E.key
3 6 2
4 E. π = D, E.key = 6, Q = B C E F
(D, F): F is in Q, w(D, F) < F.key
E F
6 F. π = D, F.key = 4, Q = F B C E

B.key = 5, C.key = 5, E.key = 6, F.key = 4

Q=FBCE
u = EXTRACT_MIN(Q) = F
A (F, E): E is in Q, w (F, E) > F.key
6 5 (F, D): D is not in Q
1 C (F, C): C is in Q, w(F, C) < F.key
B
5 5 C. π = F, C.key = 2, Q = C B E
D
3 6 2
4
E F
6

3
 11/2/2018

B.key = 5, C.key = 2, E.key = 6

Q=CBE
u = EXTRACT_MIN(Q) = C
A (C, A): A is not in Q
6 5 (C, D): D is not in Q
1 C (C, F): F is not in Q
B
5 5
D
3 6 2
4
E F
6

B.key = 5, C.key = 2, E.key = 6

Q=BE
u = EXTRACT_MIN(Q) = B
A (B, A): A is not in Q
6 5 (B, D): D is not in Q
1 C (B, E): E is in Q, w(B,E)<E.key
B
5 5 E. π = B, E.key = 3
D
3 6 2
4
E F
6

4
 11/2/2018

Prim’s algorithm example: r = 1

A A.π = NIL
6 5 B.π = D A
1 C
C.π = F
B
5 5 D.π = A 1 C
D E.π = B B
5
3 6 2
4 F.π = D D
3 2
E F
4
6
E F

Cycle property
Claim: Assume that all edge costs are distinct. Let C be any cycle in
G = (V, E), and let edge e = (v, w) be the largest edge in C. Then e is not
in any minimum spanning tree of G.
Which of the following is NOT a good approach for a proof of this
claim?
A. Assume e is in some minimum spanning tree of G, show that this
leads to a contradiction.
B. Let T be a spanning tree containing e, show that T is not a minimum
spanning tree.
C. Assume e is not in any minimum spanning tree of G. Show that your
assumption is consistent with the rest of the facts.
D. All of the above approaches are good.

5
 11/2/2018

Proof A:
Let T be a spanning tree that contains e = (v, w). We need to show that T
does not have the minimum possible weight.
Let’s delete edge e from T. This partitions the vertices of G into two
components S and V – S. Let node v be in S and node w be in V-S.
Since G has a cycle C and edge e is part of that cycle, there is a path from v to
w in G that does not involve edge e. This path will have an edge e’ with one
vertex in component S and another vertex in component V-S.
Now consider the set of edges T’ = T-{e} U {e’}. The graph (V, T’) is connected
and has no cycles, so it is a spanning tree of G. Since the weight of e > weight
e’, the weight of T’ is less than the weight of T. Therefore T is not a minimum
spanning tree.

Proof B (similar to proof A)

Assume e = (v, w) is in a minimum spanning tree T of G. (We need to show
that there is another spanning tree of G with smaller weight than T, thus
leading to a contradiction. )
Let’s delete edge e from T. This partitions the vertices of G into two
components S and V – S. Let node v be in S and node w be in V-S.
Since G has a cycle C and edge e is part of that cycle, there is a path from v to
w in G that does not involve edge e. This path will have an edge e’ with one
vertex in component S and another vertex in component V-S.
Now consider the set of edges T’ = T-{e} U {e’}. The graph (V, T’) is connected
and has no cycles, so it is a spanning tree of G. Since the weight of e > weight
e’, the weight of T’ is less than the weight of T. Therefore T is not a minimum
spanning tree. This is a contradiction. This e cannot be in a minimum
spanning tree of G

6
 11/2/2018

More on proofs
• Once a property of fact is established, it can be used in later proofs
• Example: Suppose we have a situation where we have a cycle and an
edge e in the cycle with the largest weight. We can safely claim that
that edge e will not be in the MST of a graph, by the cycle property.
• Assume, without loss of generality, that all edge weights are distinct.
• Homework 4 – proof practice

Given a connected graph G, with distinct edge

weights, let n be the number of vertices in G and
m be the number of edges. A particular edge e =
(v, w) of G is specified. Give an algorithm with
running time O(m+n) to decide whether e is
contained in a minimum spanning tree of G
1. Construct G’ from G by deleting edges with weight greater than the
weight of e. Delete e as well.
2. See if there is a path from v to w in G’, then e is not included in any
MST of G. Otherwise, e is included in the MST of G.