Chapter 3 Network Theory

3-1 Network and Flow Network: A simple, weighted, directed graph satisfies: (a) The source has no incoming edges. (b) The sink has no outgoing edges, (c) The weight Cij of the directed edge (i,j) is a nonnegative number. Cij is called the capacity of (i,j). Flow: A flow Fij of the directed edge (i,j) is a nonnegative number and satisfies: (a) Fij Cij. (b)

F
i

ij

= F ji for each vertex j, neither the source and the sink..

Eg. A network with edges label by capacity (left) and flow (right).

Eg. Fill in the missing edge flows of the left network. (Sol.) Fbc=Fab=3, Fad=Fde=2 Fce=Fez -Fde=3-2=1, Fcz= Fbc-Fce=3-1=2

Eg. Fill in the missing edge flows of the left network. (Sol.) Fbd=FabFbc=2, Fec=Fcz-Fbc=2, Fad= Fde-Fbd=1 Fez= Fde-Fec=1

Theorem The flow out of the source equals the flow into the sink. That is,

F
i

ai

= Fiz .
i

Value of the flow: The value of

F
i

ai

= Fiz .
i

Supersource and supersink: To be added in the original network without the source and the sink.

Properly oriented path and improperly oriented path:

Theorem Let P be a path from a to z in a network G. Let =min(Cij - Fij for properly oriented edges (i,j), Fij for improperly oriented edges (i,j)). Define Fij , if (i, j ) is not in P Fij * = Fij + , if (i, j ) is properly oriented in P , and then F*>F. F , if (i, j ) is improperly oriented in P ij Eg. Increase the flow of each edge in the left path. (Sol.) =min(3-1,1,3-2,5-1)=1 New flows: 1+1=2, 1-1=0, 2+1=3, 1+1=2 We have the new flows in the path:

Eg. Increase the flow of each edge in the left path. (Sol.) =min(5-1,5-2,2,6-3)=2 New flows: 1+2=3, 2+2=4, 2-2=0, 3+2=5. We have the new flows in the path:

Maximal flow algorithm

Eg. Find the maximum flow of the left path. The capacity Cij of each edge (i,j) has been labeled on the network.

(Sol.)

Eg. Find the maximum flow of the left path. The capacity Cij of each edge (i,j) has been labeled on the network.

(Sol.)

Eg. Find the maximum flow of the left path. The capacity Cij of each edge (i,j) has been labeled on the network.

(Sol.)

Cut (P, P ): A cut (P, P ) in G consists of a set P of vertices and the complement P of P, with a P and z P .

Eg. A cut (P, P ) in the left network, where P={a,b,d} and P ={c,e,f,z}.

Eg. A cut (P, P ) in the left network, where P={a,b,d} and P ={c,e,z}.

Capacity of the cut (P, P ), C(P, P ): C(P, P )= C ij iP

Eg. Find the capacity of the cut (P, P ) in the left network. (Sol.) Cbc+Cde=4+4=8

Eg. Find the capacity of the cut (P, P ) in the left network. (Sol.) Cbc+Cdc+Cde=2+2+2=6

Theorem

C
iP jP

ij

Fai . i

Max flow and min cut Theorem If equality holds in

C
iP jP

ij

Fai , then the i

flow is maximal and the cut is minimal. Moreover, equality holds in

C
iP jP

ij

Fai if and only if (a) F =C for i P and j ij ij P and (b) Fij=0 for i P i

and j P .

3-2 Matching Matching: Let G be a directed, bipartite graph with disjoint vertex set V and W in which the edges are directed from vertices in V to vertices in W. A matching for G is a set of edges E with no vertices in common. Maximal matching: A matching contains the maximum number of edges. Complete matching: A matching having the property that if v V, then (v,w) E for some w W. Eg. Two examples of matching. The black lines show maximal matching in each graph.

Maximal matching:

Maximal matching:

Matching network: Introducing a super source a and edges of capacity 1 from a to each of vi V, a super sink z and edges of capacity 1 from each of wj W to z.

Eg. Transform the left matching for G into a matching network.

(Sol.)

Eg. Find the maximal matching for the left graph. (Sol.) Matching network:

Eg. Find the maximal matching for the left graph. (Sol.) Matching network:

Eg. Find the maximal matching for the left graph. (Sol.) Matching network:

Theorem Let G be a directed, bipartite graph with disjoint vertex set V and W in which the edges are directed from vertices in V to vertices in W. (a) A flow in the matching network gives a matching in G. The vertex v V is matched with the vertex w W if and only if the flow in edge (v,w)=1. (b) A maximal flow corresponds to a maximal matching. (c) A flow whose value is |V| corresponds to a complete matching.

Halls Marriage Theorem Let G be a directed, bipartite graph with disadjoint vertex set V and W in which the edges are directed from vertices in V to vertices in W. There exists a complete matching in G if and only if |S| |R(S)| for all S V, where R(S)={w W |v S and (v,w) is an edge in G}.

Eg. There are 3 boys: a( ), b( ), c( ) and 4 girls: r( ), s( ), t( ), u( ). If a likes r and s, b likes s and u, c likes r, t and u, can each boy marry a compatible girl? (Sol.) Choose S1={a,b,c}, we have R(S1)={r,s,t,u} and |S1| =3<4=|R(S1)| Choose S2={a,b}, we have R(S2)={r,s,u} and |S2| =2<3=|R(S2)| Choose S3={a,c}, we have R(S3)={r,s,t,u} and |S3| =2<4=|R(S3)|. Choose S4={b,c}, we have R(S4)= {r,s,t,u} and |S4|=2<4=|R(S4)|, Yes! Each boy can marry a compatible girl. Eg. There are 4 members in female F4: A(Amy), B(Fanny), C(Tiffany), and D(Stacy), who choose J1J5. Let S={A,B,D}, we have R(S)={J2,J5} and |S| =3>2=|R(S)|, there is not a complete matching for the graph.

Eg. There are 3 boys: a( ), b( ), c( ) and 4 girls: r( ), s( ), t( ), u( ). If a likes r and t, b likes only t, c likes r and t, can each boy marry a compatible girl? If s() and u() are replaced by and , how do you think about it? (Sol.) Choose S={a,b,c}, we have R(S)={r,t} and |S|=3>2=| R(S)|, No! Some boy can not marry a compatible girl. For example, if a married r and b marries t, c can not marry his compatible girl. Similarly, if a married t and c married r, b can not marry his compatible girl. In case c married r and b marries t, a can not marry his compatible girl.