Illustration of the Ford-Fulkerson Method

Graph G = (V , E, c):
3/4
2 4 |f | = 18
2/4
/ 10 7/
8

⇠
8/ 10
10 0/2 8/ 6/6

⇠10
8

9/
⇠ ⇠
8/10 8/9 10/10
s 3 5 t
9/10 9/9

Residual Graph Gf = (V , Ef , cf ):
2
2 2 4

8
10

8
2 6 2
8 8
10
s 2 3 1 5 t

6.6: Maximum flow

6.6: Maximum flow T.S. 9

Frank Stajano Thomas Sauerwald

Lent 2015
Outline

Recap

Max-Flow Min-Cut Theorem

Analysis of Ford-Fulkerson

6.6: Maximum flow T.S. 2

History of the Maximum Flow Problem [Harris, Ross (1955)]

za
SECRET

3
a)
Maximum Flow is 163,000 tons per day!

On,
LL,,
cCfl

(a
_Z

oS

70

21
Z

0!
E
w 0>a

0
a_
R

-C
=
Of
~0a
a cc

'ma

In

-05
0

0
7-

4'

6.6: Maximum flow T.S. 3

Multiple Sources and Multiple Sinks (Figure 26.1)
26.1 Flow networks 713

s1 s1

10 10

12 12
3 3

∞
s2 t1 s2 t1
15 15
5 5

∞
∞
8 8
6 ∞ 6 ∞
s3 t2 s s3 t2 t
20 20
14 14

7 7

∞
13 13
s4 t3 s4 t3
18 18

∞
11 11

2 2
s5 s5
(a) (b)

Figure 26.3 Converting a multiple-source, multiple-sink maximum-flow problem into a problem

with a single source and a single sink. (a) A flow network with five sources S D fs1 ; s2 ; s3 ; s4 ; s5 g
and three sinks T D ft1 ; t2 ; t3 g. (b) An equivalent single-source, single-sink flow network. We add
a supersource s and an edge with infinite capacity from s to each of the multiple sources. We also
add a supersink t and an edge with infinite capacity from each of the multiple sinks to t.

26.1-2
Extend the flow properties and definitions to theT.S.
6.6: Maximum flow
multiple-source, multiple-sink
4
problem. Show that any flow in a multiple-source, multiple-sink flow network
Multiple Sources and Multiple Sinks (Figure 26.1)
26.1 Flow networks 713

s1 s1

10 10

12 12
3 3

∞
s2 t1 s2 t1
15 15
5 5

∞
∞
8 8
6 ∞ 6 ∞
s3 t2 s s3 t2 t
20 20
14 14

7 7

∞
13 13
s4 t3 s4 t3
18 18

∞
11 11

2 2
s5 s5
(a) (b)

Figure 26.3 Converting a multiple-source, multiple-sink maximum-flow problem into a problem

with a single source and a single sink. (a) A flow network with five sources S D fs1 ; s2 ; s3 ; s4 ; s5 g
and three sinks T D ft1 ; t2 ; t3 g. (b) An equivalent single-source, single-sink flow network. We add
a supersource s and an edge with infinite capacity from s to each of the multiple sources. We also
add a supersink t and an edge with infinite capacity from each of the multiple sinks to t.

26.1-2
Extend the flow properties and definitions to theT.S.
6.6: Maximum flow
multiple-source, multiple-sink
4
problem. Show that any flow in a multiple-source, multiple-sink flow network
Residual Graph

Original Edge
Edge e = (u, v ) ∈ E
Graph G:
flow f (u, v ) and capacity c(u, v )
6/17
u v
Residual Capacity

c(u, v ) − f (u, v )
 if (u, v ) ∈ E ,
cf (u, v ) = f (v , u) if (v , u) ∈ E ,
0 otherwise.



Residual Gf :
11
Residual Graph
u v
Gf = (V , Ef , cf ), Ef := {(u, v ) : cf (u, v ) > 0} 6

6.6: Maximum flow T.S. 5

Residual Graph with anti-parallel edges

Original Edge
Graph G:
Edge e = (u, v ) ∈ E (& possibly e0 = (v , u) ∈ E)
flow f (u, v ) and capacity c(u, v )
6/17
u v
Residual Capacity 2/4

For every pair (u, v ) ∈ V × V ,

cf (u, v ) = c(u, v ) − f (u, v ).

Residual Gf :
Residual Graph
13
Gf = (V , Ef , cf ), Ef := {(u, v ) : cf (u, v ) > 0}
u v
8

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
5/1 2 4
0
s 4/1 1/6 3/4
Flow network G 5/8 3
3/7 t
4/7

3
1/
2/3
5
13
1
11 2 4
1
11 5 1
s 3
Residual Graph Gf 7 3 4
3 t
8
2
1
2 5
1

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
5/1 2 4
0
s 4/1 1/6 3/4
Flow network G 5/8 3
3/7 t
4/7

3
1/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
5/1 2 4
0
s 4/1 1/6 2/3
Flow network G 5/8 3
2/6 t
4/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
5/1 2 4
0
s 4/1 1/6 2/3
Flow network G 5/8 3
2/6 t
4/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
5/1 2 4
0
4/1 1/6 2/3
s
Flow network G 5/8 3
2/6 t
4/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
1/1 2 4
0
0/1 1/6 2/3
s
Flow network G 5/8 3
2/6 t
4/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 5/8 3
2/6 t
4/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 5/8 3
4/7 2/6 t

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 1/8 3
0/7 2/6 t

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 1/8 3
2/6 t
0/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)
1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 1/8 3
2/6 t
0/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)
1/14
2
1/1 2 4
0
s 0/1 1/6 2/3
Flow network G 1/8 3
2/6 t
0/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)
0/14
2
0/1 2 4
0
s 0/1 0/6 2/3
Flow network G 0/8 3
2/6 t
0/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

0/14
2
0/1 2 4
0
s 0/1 0/6 2/3
Flow network G 0/8 3
2/6 t
0/7

2
0/
2/3
5

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

0/14
2
0/1 2 4
0
s 0/1 0/6 2/3
Flow network G 0/8 3
2/6 t
0/7

2
0/
2/3
5

By successively eliminating cycles we can sim-

plify and reduce the “transportation” cost of a flow.

6.6: Maximum flow T.S. 5

Example of a Residual Graph (Handout)

0/14
2
0/1 2 4
0
s 0/1 0/6 2/3
Flow network G 0/8 3
2/6 t
0/7

2
0/
2/3
5

Comment on anti-parallel edges:

You should know about this possibility
In the following proofs we will disallow anti-parallel edges
(for convenience)

6.6: Maximum flow T.S. 5

The Ford-Fulkerson Method (“Enhanced Greedy”)

0: def fordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

Augmenting path: Path

from source to sink in Gf

Questions:
How to find an augmenting path? X
Does this method terminate?
If it terminates, how good is the solution?

6.6: Maximum flow T.S. 6

Outline

Recap

Max-Flow Min-Cut Theorem

Analysis of Ford-Fulkerson

6.6: Maximum flow T.S. 10

From Flows to Cuts
Cut
A cut (S, T ) is a partition of V into S and T = V \ S such that s ∈ S
and t ∈ T .

Graph G = (V , E, c):
4
2 4

10
10

2 6

10 9 10
s 3 5 t

6.6: Maximum flow T.S. 11

From Flows to Cuts
Cut
A cut (S, T ) is a partition of V into S and T = V \ S such that s ∈ S
and t ∈ T .

Graph G = (V , E, c):
4
2 4

10
10

2 6

10 9 10
s 3 5 t

6.6: Maximum flow T.S. 11

From Flows to Cuts
Cut
A cut (S, T ) is a partition of V into S and T = V \ S such that s ∈ S
and t ∈ T .
The capacity of a cut (S, T ) is the sum of capacities of the edges
from S to T :
X X
c(S, T ) = c(u, v ) = c(u, v )
u∈S,v ∈T (u,v )∈E(S,T )

Graph G = (V , E, c):
4
2 4

10
10

2 6

10 9 10
s 3 5 t

c({s, 3}, {2, 4, 5, t}) =

6.6: Maximum flow T.S. 11
From Flows to Cuts
Cut
A cut (S, T ) is a partition of V into S and T = V \ S such that s ∈ S
and t ∈ T .
The capacity of a cut (S, T ) is the sum of capacities of the edges
from S to T :
X X
c(S, T ) = c(u, v ) = c(u, v )
u∈S,v ∈T (u,v )∈E(S,T )

Graph G = (V , E, c):
4
2 4

10
10

2 6

10 9 10
s 3 5 t

c({s, 3}, {2, 4, 5, t}) = 10 + 9 = 19

6.6: Maximum flow T.S. 11
From Flows to Cuts
Cut
A cut (S, T ) is a partition of V into S and T = V \ S such that s ∈ S
and t ∈ T .
The capacity of a cut (S, T ) is the sum of capacities of the edges
from S to T :
X X
c(S, T ) = c(u, v ) = c(u, v )
u∈S,v ∈T (u,v )∈E(S,T )

A mininum cut of a network is a cut whose capacity is minimum over

all cuts of the network.

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

10 − 2 + 8 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 11

From Flows to Cuts
Flow Value Lemma (Lemma 26.4)
Let f be a flow with source s and sink t, and let (S, T ) be any cut of G.
Then the value of the flow is equal to the net flow across the cut, i.e.,
X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts

X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts

X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

|f | =

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts

X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

X
|f | = f (s, w)
w∈V

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts

X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

X X X X 
|f | = f (s, w) = f (u, w) − f (w, u)
w∈V u∈S (u,w)∈E (w,u)∈E

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
From Flows to Cuts

X X
|f | = f (u, v ) − f (v , u).
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

X X X X 
|f | = f (s, w) = f (u, w) − f (w, u)
w∈V u∈S (u,w)∈E (w,u)∈E
X X
= f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

Graph G = (V , E, c): |f | = 16
0/4
2 4
8/
0

6/
/1

2/2 6/6

10
10

s 3 5 t
6/10 8/9 10/10

8 + 8 − 6 + 6 = 16
6.6: Maximum flow T.S. 11
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut.

6.6: Maximum flow T.S. 12

Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

6.6: Maximum flow T.S. 12

Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

0/4
2 4
0

6/
8/
/1

2/2 6/6
10
8
10

s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

|f | =

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)
X
≤ f (u, v )
(u,v )∈E(S,T )

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)
X
≤ f (u, v )
(u,v )∈E(S,T )
X
≤ c(u, v )
(u,v )∈E(S,T )

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Weak Duality betwen Flows and Cuts
Weak Duality (Corollary 26.5)
Let f be any flow and (S, T ) be any cut. Then the value of f is bounded
from above by the capacity of the cut (S, T ), i.e.,

|f | ≤ c(S, T ).

X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)
X
≤ f (u, v )
(u,v )∈E(S,T )
X
≤ c(u, v ) = c(S, T ).
(u,v )∈E(S,T )

0/4 4
2 4 2 4
0

6/
8/
/1

10

10
2/2 6/6 2 6

8
10
8
10

10 9 10
s 3 5 t s 3 5 t
6/10 8/9 10/10
|f | = 10 − 2 + 8 = 16 c(S, T )=10+9=19
6.6: Maximum flow T.S. 12
Max-Flow Min-Cut Theorem

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

6.6: Maximum flow T.S. 13

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 3 ⇒ 1:

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 3 ⇒ 1:
Suppose that (S, T ) is a cut with c(S, T ) = |f |

3/4 4
G 2 4 Gf 2 4
10 10
0/
9/
7/

10
0/2 6/6 2 6

8
10

1
8

10 9 10
s 3 5 t s 3 5 t
9/10 9/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 3 ⇒ 1:
Suppose that (S, T ) is a cut with c(S, T ) = |f |
By Corollary 26.5, for any flow ef ,

3/4 4
G 2 4 Gf 2 4
10 10
0/
9/
7/

10
0/2 6/6 2 6

8
10

1
8

10 9 10
s 3 5 t s 3 5 t
9/10 9/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 3 ⇒ 1:
Suppose that (S, T ) is a cut with c(S, T ) = |f |
By Corollary 26.5, for any flow ef , |ef | ≤ c(S, T ) = |f |.

3/4 4
G 2 4 Gf 2 4
10 10
0/
9/
7/

10
0/2 6/6 2 6

8
10

1
8

10 9 10
s 3 5 t s 3 5 t
9/10 9/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 3 ⇒ 1:
Suppose that (S, T ) is a cut with c(S, T ) = |f |
By Corollary 26.5, for any flow ef , |ef | ≤ c(S, T ) = |f |.
Hence f is a maximum flow.

3/4 4
G 2 4 Gf 2 4
10 10
0/
9/
7/

10
0/2 6/6 2 6

8
10

1
8

10 9 10
s 3 5 t s 3 5 t
9/10 9/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 1 ⇒ 2:

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 1 ⇒ 2:
For the sake of contradicion, suppose there is an augmenting path with
respect to f .

0/4 4
G 2 4 Gf 2 4
10
0/ 4
6/

6
8/

10
2/2 6/6 2 6

8
10

1
8

6 8
10
s 3 5 t s 4 3 1 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 1 ⇒ 2:
For the sake of contradicion, suppose there is an augmenting path with
respect to f .
Then we can improve f by increasing the flow along this path.

0/4 4
G 2 4 Gf 2 4
10
0/ 4
6/

6
8/

10
2/2 6/6 2 6

8
10

1
8

6 8
10
s 3 5 t s 4 3 1 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 1 ⇒ 2:
For the sake of contradicion, suppose there is an augmenting path with
respect to f .
Then we can improve f by increasing the flow along this path.
Hence f cannot be a maximum flow.

0/4 4
G 2 4 Gf 2 4
10
0/ 4
6/

6
8/

10
2/2 6/6 2 6

8
10

1
8

6 8
10
s 3 5 t s 4 3 1 5 t
6/10 8/9 10/10

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.

3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S

3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.

3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T )

3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).

3
3/4
G 2 4 Gf 2 1 4

/1
0 7/
8
2 1 6 1
9/

9
10 /10 0/2
10

 6/6
10

9 9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).

3
3/4
G 2 4 Gf 2 1 4

/1
0 7/
8
2 1 6 1
9/

9
10 /10 0/2
10

 6/6
10

9 9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).

3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S)
3
3/4
G 2 4 Gf 2 1 4
10 7/
0/ 8
2 1 6 1
9/

9
10
0/2 6/6
10

1
9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.
3
3/4
G 2 4 Gf 2 1 4
0 7
0/
1
 /8 2 1 6 1
9/

9
1 1/2 
10
0/2 6/6
10

9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.
3
3/4
G 2 4 Gf 2 1 4
0 7
0/
1
 /8 2 1 6 1
9/

9
1 1/2 
10
0/2 6/6
10

9
9 10
s 3 5 t s 1 3 5 t
9/10 9/9 10/10
6.6: Maximum flow T.S. 14
Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.

|f | =

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.
X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.
X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)
X
= c(u, v )
(u,v )∈E(S,T )

6.6: Maximum flow T.S. 14

Key Lemma

Key Lemma (Theorem 26.6)

The following three conditions are all equivalent for any flow f :
1. f is a maximum flow
2. There is no augmenting path in Gf
3. There exists a cut (S, T ) such that c(S, T ) = |f |

Proof 2 ⇒ 3:
Let f be a flow with no augmenting paths.
Let S be the nodes reachable from s in Gf , T := V \ S ⇒ s ∈ S, t 6∈ S.
(u, v ) ∈ E(S, T ) ⇒ f (u, v ) = c(u, v ).
(v , u) ∈ E(T , S) ⇒ f (v , u) = 0.
X X
|f | = f (u, v ) − f (v , u)
(u,v )∈E(S,T ) (v ,u)∈E(T ,S)
X
= c(u, v ) = c(S, T )
(u,v )∈E(S,T )

6.6: Maximum flow T.S. 14

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≤”:

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≤”:
For any flow f and cut (S, T ), |f | ≤ c(S, T ) (Corollary 26.5)

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≤”:
For any flow f and cut (S, T ), |f | ≤ c(S, T ) (Corollary 26.5)
⇒
max |f | ≤ min c(S, T )
f S,T ⊆V

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≥”:
Let fmax be a maximum flow

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≥”:
Let fmax be a maximum flow
Key Lemma ⇒ there is a cut (S, T ) with c(S, T ) = |fmax |.

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Theorem
The value of the max-flow is equal to the capacity of the min-cut, that is

max |f | = min c(S, T ).

f S,T ⊆V

Proof of “≥”:
Let fmax be a maximum flow
Key Lemma ⇒ there is a cut (S, T ) with c(S, T ) = |fmax |.
⇒
max |f | = |fmax | = c(S, T ) ≥ min c(S, T )
f S,T ⊆V

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Max-Flow Min-Cut Theorem algorithm for computing min-cut:

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Max-Flow Min-Cut Theorem algorithm for computing min-cut:

Run Ford-Fulkerson until it terminates with a flow fmax

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Max-Flow Min-Cut Theorem algorithm for computing min-cut:

Run Ford-Fulkerson until it terminates with a flow fmax
Define S as the set of nodes reachable from s in Gf , and T = V \ S

6.6: Maximum flow T.S. 15

Proof of the Max-Flow Min-Cut Theorem
Key Lemma
The following conditions are equivalent for any flow f :
1. f is a maximum flow.
2. There is no augmenting path in Gf .
3. There exists a cut (S, T ) such that c(S, T ) = |f |.

Max-Flow Min-Cut Theorem algorithm for computing min-cut:

Run Ford-Fulkerson until it terminates with a flow fmax
Define S as the set of nodes reachable from s in Gf , and T = V \ S
⇒ (S, T ) is a minimal (s, t)-cut whose capacity is equal to |fmax |

6.6: Maximum flow T.S. 15

Outline

Recap

Max-Flow Min-Cut Theorem

Analysis of Ford-Fulkerson

6.6: Maximum flow T.S. 16

Analysis of Ford-Fulkerson

0: def FordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

6.6: Maximum flow T.S. 17

Analysis of Ford-Fulkerson

0: def FordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

Lemma
If all capacities c(u, v ) are integral, then the flow at every iteration of
Ford-Fulkerson is integral.

6.6: Maximum flow T.S. 17

Analysis of Ford-Fulkerson

0: def FordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

Lemma
If all capacities c(u, v ) are integral, then the flow at every iteration of
Ford-Fulkerson is integral.

Flow before iteration integral

& capacities in Gf are integral
⇒ Flow after iteration integral

6.6: Maximum flow T.S. 17

Analysis of Ford-Fulkerson

0: def FordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

Lemma
If all capacities c(u, v ) are integral, then the flow at every iteration of
Ford-Fulkerson is integral.

Theorem
For integral capacities c(u, v ), Ford-Fulkerson terminates after V · C
iterations, where C := maxu,v c(u, v ) and returns the maximum flow.

6.6: Maximum flow T.S. 17

Analysis of Ford-Fulkerson

0: def FordFulkerson(G)
1: initialize flow to 0 on all edges
2: while an augmenting path in Gf can be found:
3: push as much extra flow as possible through it

Lemma
If all capacities c(u, v ) are integral, then the flow at every iteration of
Ford-Fulkerson is integral.

Theorem
For integral capacities c(u, v ), Ford-Fulkerson terminates after V · C
iterations, where C := maxu,v c(u, v ) and returns the maximum flow.

at the time of termination, no augmenting path

⇒ Ford-Fulkerson returns maxflow (Key Lemma)

6.6: Maximum flow T.S. 17

Slow Convergence of Ford-Fulkerson (Figure 26.7)

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1
000 000
0/1
s 0/1 t
0/1
000 000
v 0/1

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1 u
000 000 0 100
0/1 100 0

s 0/1 t s 1 t
0/1 100
000 000 0 100
0
v 0/1 v

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1 u
000 000 0 100
0/1 100 0

s 0/1 t s 1 t
0/1 100
000 000 0 100
0
v 0/1 v

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1 u
000 000 0 100
1/1 100 0

s 1/1 t s 1 t
0/1 100
000 000 0 100
0
v 1/1 v

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1 u
000 000 100
1/1 999 0
1
s 1/1 t s 1 t
0/1 100
000 000 0 999
v 1/1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 0/1 u
000 000 100
1/1 999 0
1
s 1/1 t s 1 t
0/1 100
000 000 0 999
v 1/1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 100
1/1 999 0
1
s 0/1 t s 1 t
1/1 100
000 000 0 999
v 1/1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 999
1/1 999
1 1
s 0/1 t s 1 t
1/1
000 000 999
999
v 1/1 1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 999
1/1 999
1 1
s 0/1 t s 1 t
1/1
000 000 999
999
v 1/1 1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 999
2/1 999
1 1
s 1/1 t s 1 t
1/1
000 000 999
999
v 2/1 1 v 1

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 999
2/1 998
2 1
s 1/1 t s 1 t
1/1
000 000 999
998
v 2/1 1 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 1/1 u
000 000 999
2/1 998
2 1
s 1/1 t s 1 t
1/1
000 000 999
998
v 2/1 1 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 999
2/1 998
2 1
s 0/1 t s 1 t
2/1
000 000 999
998
v 2/1 1 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 998
2/1 998
2 2
s 0/1 t s 1 t
2/1
000 000 998
998
v 2/1 2 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 998
2/1 998
2 2
s 0/1 t s 1 t
2/1
000 000 998
998
v 2/1 2 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 998
3/1 998
2 2
s 1/1 t s 1 t
2/1
000 000 998
998
v 3/1 2 v 2

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 998
3/1 997
3 2
s 1/1 t s 1 t
2/1
000 000 998
997
v 3/1 2 v 3

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 2/1 u
000 000 998
3/1 997
3 2
s 1/1 t s 1 t
2/1
000 000 998
997
v 3/1 2 v 3

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 3/1 u
000 000 998
3/1 997
3 2
s 0/1 t s 1 t
3/1
000 000 998
997
v 3/1 2 v 3

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 3/1 u
000 000 997
3/1 997
3 3
s 0/1 t s 1 t
3/1
000 000 997
997
v 3/1 3 v 3

G Gf

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 3/1 u
000 000 997
3/1 997
3 3
s 0/1 t s 1 t
3/1
000 000 997
997
v 3/1 3 v 3

G Gf

Number of iterations is at least C := maxu,v c(u, v )!

6.6: Maximum flow T.S. 18

Slow Convergence of Ford-Fulkerson (Figure 26.7)

u 3/1 u
000 000 997
3/1 997
3 3
s 0/1 t s 1 t
3/1
000 000 997
997
v 3/1 3 v 3

G Gf

Number of iterations is at least C := maxu,v c(u, v )!

For irrational capacities, Ford-Fulkerson

may even fail to terminate!

6.6: Maximum flow T.S. 18

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2 3 4 5

6.6: Maximum flow T.S. 19

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2 3 4 5

./∞
./∞ ./∞
t

6.6: Maximum flow T.S. 19

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 0/1
2 3 4 5

./∞
./∞ ./∞
t

6.6: Maximum flow T.S. 19

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 0/1 0/φ
2 3 4 5
√
./∞ φ2 = 1 −./∞
φ, φ = ( 5./∞
− 1)/2 ≈ 0.618

6.6: Maximum flow T.S. 19

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 0/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

6.6: Maximum flow T.S. 19

Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 0/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 0/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 1, |f | = 0

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 
0/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 1, |f | = 0

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 1, |f | = 1

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 1, |f | = 1

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

0/1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 2, |f | = 1

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞


0/1 
1/1 
0/φ

2 3 4 5

./∞
./∞ ./∞
t

Iteration: 2, |f | = 1

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1 − φ/1 φ/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 2, |f | = 1 + φ

1 1 φ
2 3 4 5

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1 − φ/1 φ/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 2, |f | = 1 + φ

1−φ φ
φ
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1 − φ/1 φ/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 3, |f | = 1 + φ

1−φ φ
φ
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞


φ/1
1 − φ /1 2

1−

φ/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 3, |f | = 1 + φ

1−φ φ
φ
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 3, |f | = 1 + 2 · φ

1−φ φ
φ
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 3, |f | = 1 + 2 · φ

1−φ
1 φ
2 3 4 5
φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1/1 0/φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 4, |f | = 1 + 2 · φ

1−φ
1 φ
2 3 4 5
φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
φ

1− /1 
1/1 
0/φ

2 3 4 5

./∞
./∞ ./∞
t

Iteration: 4, |f | = 1 + 2 · φ

1−φ
1 φ
2 3 4 5
φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

1/1 φ/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 4, |f | = 1 + 2 · φ + φ2

1−φ
1 φ
2 3 4 5
φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

1/1 φ/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 4, |f | = 1 + 2 · φ + φ2

1−φ φ3
1
2 3 4 5
φ φ − φ3

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

1/1 φ/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 5, |f | = 1 + 2 · φ + φ2

1−φ φ3
1
2 3 4 5
φ φ − φ3

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞


1/1  φ − φ3 /φ
φ/1
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 5, |f | = 1 + 2 · φ + φ2

1−φ φ3
1
2 3 4 5
φ φ − φ3

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 5, |f | = 1 + 2 · φ + 2 · φ2

1−φ φ3
1
2 3 4 5
φ φ − φ3

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

2
1 − φ /1 1/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

Iteration: 5, |f | = 1 + 2 · φ + 2 · φ2

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 1/1 φ − φ3 /φ
2 3 4 5

./∞
./∞ ./∞
t

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5

./∞
./∞ ./∞
t

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞ ./∞
t

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

More generally, s

1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

More generally, s
1−φ2i 1 φ−φ2i+1
For every i = 0, 1, . . . after iteration 1 + 4 · i: −→ , −→, ←−
1−φ φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

More generally, s
1−φ2i 1 φ−φ2i+1
For every i = 0, 1, . . . after iteration 1 + 4 · i: −→ , −→, ←−
1 − φ does not terminate!
Ford-Fulkerson φ3
1
2 3 4 5
φ 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

More generally, s
1−φ2i 1 φ−φ2i+1
For every i = 0, 1, . . . after iteration 1 + 4 · i: −→ , −→, ←−
1 − φ does not terminate!
Ford-Fulkerson φ3
1
2 3 4 5
|f | = 1 + 2 2i Φi < 7
P
φ k =1 1−φ

t
6.6: Maximum flow T.S. 19
Non-Termination of Ford-Fulkerson for Irrational Capacities
s

./∞ ./∞ ./∞

In summary:
2
1 − φ /1 0 1 0 1/1 φ − φ3 /φ
After
2 iteration 1: −→, −→,
3 ←−, |f | = 1 4 5
1−φ2 1 φ−φ3 2
After iteration 5: −→ , −→, ←− , |f |./∞
= 1 + 2φ + 2φ
./∞1−φ4 5 ./∞
1 φ−φ
After iteration 9: −→ , −→, ←− , |f | = 1 + 2φ + 2φ2 + 2φ3 + 2φ4
t

More generally, s
1−φ2i 1 φ−φ2i+1
For every i = 0, 1, . . . after iteration 1 + 4 · i: −→ , −→, ←−
1 − φ does not terminate!
Ford-Fulkerson φ3
1
2 3 4 5
|f | = 1 + 2 2i Φi < 7
P
φ k =1 1−φ
It does not even converge to a maximum flow!

t
6.6: Maximum flow T.S. 19
Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities

6.6: Maximum flow T.S. 20

Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities
Runtime: O(E · |f ∗ |) = O(E · V · C)

6.6: Maximum flow T.S. 20

Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities
Runtime: O(E · |f ∗ |) = O(E · V · C)

Capacity-Scaling Algorithm

6.6: Maximum flow T.S. 20

Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities
Runtime: O(E · |f ∗ |) = O(E · V · C)

Capacity-Scaling Algorithm
Idea: Find an augmenting path with high capacity
Consider subgraph of Gf consisting of edges (u, v ) with cf (u, v ) > ∆
scaling parameter ∆, which is initially 2dlog2 Ce and 1 after termination
Runtime: O(E 2 · log C)

6.6: Maximum flow T.S. 20

Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities
Runtime: O(E · |f ∗ |) = O(E · V · C)

Capacity-Scaling Algorithm
Idea: Find an augmenting path with high capacity
Consider subgraph of Gf consisting of edges (u, v ) with cf (u, v ) > ∆
scaling parameter ∆, which is initially 2dlog2 Ce and 1 after termination
Runtime: O(E 2 · log C)

Edmonds-Karp Algorithm

6.6: Maximum flow T.S. 20

Summary and Outlook

Ford-Fulkerson Method
works only for integral (rational) capacities
Runtime: O(E · |f ∗ |) = O(E · V · C)

Capacity-Scaling Algorithm
Idea: Find an augmenting path with high capacity
Consider subgraph of Gf consisting of edges (u, v ) with cf (u, v ) > ∆
scaling parameter ∆, which is initially 2dlog2 Ce and 1 after termination
Runtime: O(E 2 · log C)

Edmonds-Karp Algorithm
Idea: Find the shortest augmenting path in Gf
Runtime: O(E 2 · V )

6.6: Maximum flow T.S. 20