Network Flow

IT307
Network flow problems

By
K V Arya
Network flow problems

- Introduction of: network, max-flow problem capacity, flow - Ford-Fulkerson method pseudo code, residual networks, augmenting paths cuts of networks max-flow min-cut theorem example of an execution
analysis, running time,

variations of the max-flow problem
Introduction network
Practical examples of a network
- liquids flowing through pipes - parts through assembly lines - current through electrical network
- information through communication network

- goods transported on the road
Introduction - network
Representation
Flow network: directed graph G=(V,E)
v1 v3 v1 v3
Example: oil pipeline
v2
v4
source
v2
v4
sink
Introduction max-flow problem

Representation
v1 v3 v1 v3
v2
v4
source
v2
v4
sink
Informal definition of the max-flow problem: What is the greatest rate at which material can be shipped from the source to the sink without violating any capacity contraints?
Introduction - capacity
Representation
3 v1 v3 v1 6 S v3
8
S 3 6
t 8
v2
6 12
v4
v2
v4
c(u,v)=12
6
Big pipe
v
c(u,v)=6
Small pipe
Introduction - capacity
Representation
3 v1 v3 v1 6 S v3
8
S 3 6
t 8
v2
v4
v2
v4
If (u,v)
v2
E
6
c(u,v) = 0
v4
0
0 v4 0 v3
Introduction flow
Representation
3 v1 v3 v1 6 S v3
8
S 3 6
t 8
v2
6 6/12
v4
v2
v4
f(u,v)=6
6/6
Flow below capacity

v
f(u,v)=6
Maximum flow
Introduction flow
Representation
3 v1 v3 v1 6 S v3
8
S 3 6
t 8
v2
v4
v2
v4
Introduction flow
Representation
3 v1 v3 v1 6 S v3
8
S 3 6/6
t 6/8
v2
6/6
v4
v2
v4
Introduction flow
Representation
3 v1 v3 v1 6 S v3
8
S 3 6/6
t 6/8
v2
6/6
v4
v2
v4
Introduction flow
Representation
3/3 v1 v3 v1 3/6 S v3
3/8
S 3 6/6
t 6/8
v2
6/6
v4
v2
v4
Introduction flow
Representation
3/3 v1 v3 v1 3/6 S v3
3/8
S 3 6/6
t 6/8
v2
6/6
v4
v2
v4
Introduction flow
Representation
3/3 v1 v3 v1 3/6 S v3
5/8
S 3 6/6
2/3
t 8/8
v2
6/6
v4
v2
v4
Introduction cancellation
Representation
3/3 v1 v3 v1 3/6 S v3
5/8
S 3 6/6
2/3
t 8/8
v2
6/6
v4
v2
v4
Introduction cancellation
Representation
3/3 v1 v3 v1 4/6 S v3
6/8
S
3/3
t 1/3 8/8
6/6
v2
5/6
v4
v2
v4
u 10 4 8/10
u 4 8/10
u 3/4 5/10
u 4
Flow properties
Flow in G = (V,E): f: V x V R with 3 properties: V : f(u,v) < c(u,v) V : f(u,v) = - f(v,u) V \ {s,t} :
v V
1) Capacity constraint: For all u,v 2) Skew symmetry: For all u,v
3) Flow conservation: For all u
f(u,v) = 0
Flow properties
Flow in G = (V,E): f: V x V R with 3 properties: V : f(u,v) < c(u,v) V : f(u,v) = - f(v,u) V \ {s,t} :
v V
1) Capacity constraint: For all u,v 2) Skew symmetry: For all u,v
3) Flow conservation: For all u

Flow network G = (V,E)
f(u,v) = 0
v1
12/12
v3
Note: by skew symmetry

t
f (v3,v1) = - 12
v2
11/14
v4
Net flow and value of a flow

u u
3/4 5/10 4
Net Flow: positive or negative value of f(u,v)
f(u,v) = 5
f(v,u) = -5
8/10
3/3 v1 v3
Value of a Flow f: Def: |f| = v V f(s,v)
6/8
S
3/3
4/6
1/3
6/6
v2
5/6
v4
8/8
The max-flow problem

Informal definition of the max-flow problem:
What is the greatest rate at which material can be shipped from the source to the sink without violating any capacity contraints?
Formal definition of the max-flow problem:
The max-flow problem is to find a valid flow for a given weighted directed graph G, that has the maximum value over all valid flows.
The Ford-Fulkerson method

a way how to find the max-flow
This method contains 3 important ideas: 1) residual networks
2) augmenting paths
3) cuts of flow networks
Ford-Fulkerson pseudo code
1 initialize flow f to 0
2 while there exits an augmenting path p

3 do augment flow f along p
4 return f
Ford Fulkerson residual networks

The residual network Gf of a given flow network G with a valid flow f consists of the same vertices v V as in G which are linked with residual edges (u,v) Ef that can admit more strictly positive net flow. The residual capacity cf represents the weight of each edge Ef and is the amount of additional net flow f(u,v) before exceeding the capacity c(u,v)
v1
residual network Gf = (V,Ef)

v1 v3
12/12
v3
v2
11/14
v4
v2
v4
cf(u,v) = c(u,v) f(u,v)
Ford Fulkerson residual networks

The residual network Gf of a given flow network G with a valid flow f consists of the same vertices v V as in G which are linked with residual edges (u,v) Ef that can admit more strictly positive net flow. The residual capacity cf represents the weight of each edge Ef and is the amount of additional net flow f(u,v) before exceeding the capacity c(u,v)
v1

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4
Ford Fulkerson augmenting paths

Definition: An augmenting path p is a simple (free of any cycle) path from s to t in the residual network Gf
Residual capacity of p
cf(p) = min{cf (u,v): (u,v) is on p}

v1

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4

Definition: An augmenting path p is a simple (free of any cycle) path from s to t in the residual network Gf
Residual capacity of p

v1

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4
Augmenting path

We define a flow: fp: V x V R such as:
cf(p) if (u,v) is on p
fp(u,v) = - cf(p) if (v,u) is on p 0 Flow network G = (V,E)
v1
otherwise residual network Gf = (V,Ef)

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4

v1

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4
Our virtual flow fp along the augmenting path p in Gf
Ford Fulkerson augmenting the flow

v1

v1
12/12
v3
12
v3
v2
11/14
v4
v2
3 11
v4
New flow:
f: V x V
R : f=f + fp
Our virtual flow fp along the augmenting path p in Gf
Ford Fulkerson new valid flow

proof of capacity constraint Lemma: f :VxV R : f = f + fp in G
cf(p) fp(u,v) = if (u,v) is on p - cf(p) if (v,u) is on p 0 otherwise
Capacity constraint:
For all u,v Proof:
V, we require f(u,v) < c(u,v)
fp (u ,v) < cf (u ,v) = c (u ,v) f (u ,v) (f + fp) (u ,v) = f (u ,v) + fp (u ,v) < c (u ,v)

proof of Skew symmetry Lemma: f :VxV R : f = f + fp in G
Skew symmetry:
For all u,v Proof: (f + fp)(u ,v) = f (u ,v) + fp (u ,v) = - f (v ,u) fp (v ,u) = - (f (v ,u) + fp (v ,u)) = - (f + fp) (v ,u) V, we require f(u,v) = - f(v,u)

proof of flow conservation Lemma: f :VxV R : f = f + fp in G
Flow conservation:
For all u Proof: u V {s ,t}
v V
V \ {s,t} :
v V
f(u,v) = 0
(f + fp) (u ,v) = v V ,v) + fp (u ,v)) (f(u
=v Vf (u ,v) + v Vfp (u ,v) = 0 + 0 = 0

Lemma: | (f + fp) | = | f | + | fp |
Value of a Flow f:
Def: Proof: | (f + fp) | = v V(f + fp) (s ,v) = v V (s ,v) + fp (s ,v)) (f |f| =
v V
f(s,v)
=
v V
f (s ,v) +
v
fp (s ,v) = | f | + | fp |

Lemma: f :VxV R : f = f + fp in G
| (f + fp) | = | f | + | fp | > | f | Lemma shows: if an augmenting path can be found then the above flow augmentation will result in a flow improvement. Question: If we cannot find any more an augmenting path is our flow then maximum? Idea: The flow in G is maximum contains no augmenting path. the residual Gf
Ford Fulkerson cuts of flow networks

New notion: cut (S,T) of a flow network
A cut (S,T) of a flow network G=(V,E) is a partiton of V into S and T = V \ S such that s S and t T.
Practical example
v1
12/12
v3
In the example: S = {s,v1,v2) , T = {v3,v4,t}
Net flow f(S ,T) = f(v1,v3) + f(v2,v4) + f(v2,v3) = 12 + 11 + (-0) = 23
v2
11/14
v4
Capacity c(S,T) = c(v1,v3) + c(v2,v4) = 12 + 14 = 26 f (u, v)

S v T
Implicit summation notation: f (S, T) =

u

Lemma:
the value of a flow in a network is the net flow across any cut of the network f (S ,T) = | f |
12/12
v1
v3
v2
11/14
v4
proof

Assumption: The value of any flow f in a flow network G is bounded from above by the capacity of any cut of G Lemma: | f | < c (S, T)
v1
| f | = f (S, T) = < f (u, v)

S
u S v T
12/12
v3
c (u, v)
v2
u S v T
= c (S, T)
Hence, | f | < c (S, T)
11/14
v4
F. Fulkerson: Max-flow min-cut theorem

If f is a flow in a flow network G = (V,E) with source s and sink t, then the following conditions are equivalent: 1. f is a maximum flow in G. 2. The residual network Gf contains no augmenting paths. 3. | f | = c (S, T) for some cut (S, T) of G.
proof: (1) (2):
We assume for the sake of contradiction that f is a maximum flow in G but that there still exists an augmenting path p in Gf. Then as we know from above, we can augment the flow in G according to the formula: f = f + fp. That would create a flow f that is strictly greater than the former flow f which is in contradiction to our assumption that f is a maximum flow.

proof: original flow network G residual network Gf
3
4/6 2 6 S 1/3 6/6 v2 5/6 v4 6 8/8 v2 1 t S 2 5 v4 8 v1 3 1 4 t v3 2
(2)
(3):
6/8
3/3
v1 3/3 v3

proof: residual network Gf
(2)
(3): Define
path p from s to v in Gf } S according to (2))
2 6 S
3
v1 3 1 2 6 v2 1 5
v3 2 4 t 8
S = {v V |
T = V \ S (note t
v4

proof: original network G
(2)
(3): Define
v1
3/3
v3 3/3 4/6 6/8 S 1/3 6/6 v2 1 5/6 v4 8/8
S = {v V |
path p from s to v in Gf } S according to (2))
T = V \ S (note t
for u S, v T: f (u, v) = c (u, v) (otherwise (u, v) Ef and v S)

| f | = f (S, T) = c (S, T)

If f is a fow in a flow network G = (V,E) with source s and sink t, then the following conditions are equivalent: 1. f is a maximum flow in G. 2. The residual network Gf contains no augmenting paths. 3. | f | = c (S, T) for some cut (S, T) of G.
proof:
(3)
(1): as proofed before | f | = f (S, T) < c (S, T)
the statement of (3) : | f | = c (S, T) implies that f is a maximum flow
The basic Ford Fulkerson algorithm

1 for each edge (u, v) E [G] 2 do f [u, v] = 0 3 f [v, u] = 0 4 while there exists a path p from s to t in the residual network Gf 5 do cf (p) = min {cf (u, v) | (u, v) is in p} 6 for each edge (u, v) in p 7 do f [u, v] = f [u, v] + cf (p) 8 f [v, u] = - f [u, v]

example of an execution
(residual) network Gf
v1
12
v3
1 2 3 4
t
v2
14
v4
5 6 7 8
for each edge (u, v) E [G] do f [u, v] = 0 f [v, u] = 0 while there exists a path p from s to t in the residual network Gf do cf(p) = min{cf(u, v) | (u, v) p} for each edge (u, v) in p do f [u, v] = f [u, v] + cf(p) f [v, u] = - f [u, v]

v1
0/12
v3
1 2 3 4
t
v2
0/14
v4
5 6 7 8

v1
12
v3
1 2 3 4
t
v2
14
v4
5 6 7 8

v1
12
v3
1 2 3 4
t
v2
14
v4
5 6 7 8
temporary variable: cf (p) = 12

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
temporary variable:
12/12
v3
cf (p) = 12
v2
14
v4

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
14
v4

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
14
v4

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
14
v4

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
12/12
v3
temporary variable:
cf (p) = 4
v2
14
v4

v1
12
v3
1 2 3 4
t
v2
14
v4
new flow network G

v1
5 6 7 8
temporary variable:
12/12
v3
cf (p) = 4
v2
4/14
v4

v1
12
v3
1 2 3 4
t
4
v2
10
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
4/14
v4

v1
12
v3
1 2 3 4
t
4
v2
10
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
4/14
v4

v1
12
v3
1 2 3 4
t
4
v2
10
v4
new flow network G

v1
5 6 7 8
12/12
v3
v2
4/14
v4

v1
12
v3
1 2 3 4
t
4
v2
10
v4
new flow network G

v1
5 6 7 8
temporary variable:
12/12
v3
cf (p) = 7
v2
4/14
v4

v1
12
v3
1 2 3 4
t
4
v2
10
v4
new flow network G

v1
5 6 7 8
temporary variable:
12/12
v3
cf (p) = 7
v2
11/14
v4

v1
12
v3
1 2 3 4
t
11
v2
v4
new flow network G

v1
5 6 7 8
for each edge (u, v) E [G] do f [u, v] = 0 f [v, u] = 0 while there exists a path p from s to t in the residual network G f do cf(p) = min{c f(u, v) | (u, v) p} for each edge (u, v) in p do f [u, v] = f [u, v] + c f(p) f [v, u] = - f [u, v]
12/12
v3
v2
11/14
v4

v1
12
v3
1 2 3 4
t
11
v2
v4
new flow network G

v1
5 6 7 8
12/12
v3
Finally we have:
S t
| f | = f (s, V) = 23
v2
11/14
v4
Analysis of the Ford Fulkerson algorithm

Running time
1 2 3 4 5 6 7 8 for each edge (u, v) E [G] do f [u, v] = 0 f [v, u] = 0 while there exists a path p from s to t in the residual network Gf do cf(p) = min{cf(u, v) | (u, v) p} for each edge (u, v) in p do f [u, v] = f [u, v] + cf(p) f [v, u] = - f [u, v]
The running time depends on how the augmenting path p in line 4 is determined.

Running time (arbitrary choice of p)
1 2 3 4 5 6 7 8 for each edge (u, v) E [G] do f [u, v] = 0 O(|E|) f [v, u] = 0 while there exists a path p from s to t O(|E|) in the residual network Gf do cf(p) = min{cf(u, v) | (u, v) p} for each edge (u, v) in p O(|E|) do f [u, v] = f [u, v] + cf(p) f [v, u] = - f [u, v]
running time: O ( |E| |fmax| )
O(|E| |fmax|)
with fmax as maximum flow

(1) The augmenting path is chosen arbitrarily and all capacities are integers

Consequencies of an arbitrarily choice: Example if |f*| is large:
v1
v2


v1 1/1
residual network Gf
v1
v2
v2


v1
residual network Gf
v1
v2
v2


Running time with Edmonds-Karp algorithm
Informal idea of the proof: (1) for all vertices v V\{s,t}:

S
u1
v3
the shortest path distance f(s,v) in Gf increases monotonically with each flow augmentation
f (s,v) < f
u2
(s,v)
v4
running time: O ( |V| |E| ) (2) Edmonds-Karp algorithm augmenting path is found by breath-first search and has to be a shortest path from p to t


S
u1
v3
. f(s,v) <
f
u2
(s,v)
v4
f(s,u2)
=1


S
Two flow augmentations later

u1
v3
. f(s,v) <
f
u2
(s,v)
v4
f(s,u2)
=3

The total number of flow augmentations performed by the algorithm is at most O(|V| |E|)
S
u1
v3
u2
v4

Def: critical edge (u,v) if cf(p) = cf(u,v)

u1
v3
At least one edge on p must be critical.

How many times can (u,v) be critical during an execution of Edmonds-Karp algorithm?
u2
v4
Critical edges on the path p

Idea: (u,v) can become critical at most O(V) times (After beeing critical it dissappears in Gf and can only reappear after net flow is decreased)
S
u1
v3
u2
v4
our example: edge (u2,v4) running time: O ( |V| |E| ) (2) Edmonds-Karp algorithm augmenting path is found by breath-first search and has to be a shortest path from p to t

Idea: (u,v) can become critical at most O(V) times (After beeing critical it dissappears in Gf and can only reappear after net flow is decreased)
S
Two flow augmentations later

u1
v3
u2
v4
our example: edge (u2,v4) has disappeared

Idea: (u,v) can become critical at most O(V) times (After beeing critical it disappears in Gf and can only reappear after net flow is decreased)
S
After the third flow augmentation

u1
v3
2
u2 v4
our example: edge (u2,v4) reappeared but is now unreachable from s running time: O ( |V| |E| ) (2) Edmonds-Karp algorithm augmenting path is found by breath-first search and has to be a shortest path from p to t

Idea:
(u,v) can become critical at most O(|V|) times

O(|E|) pairs of vertices exist in Gf
All critical edges in O(|V| |E|)

O(|E|) for breath first search for p total running time O(|V| |E|) running time: O ( |V| |E| )
(2) Edmonds-Karp algorithm

In Ford Fulkerson algorithm
If augmenting path is found by breath-first search and has to be a shortest path from p to t then it become E-K algorithm
Different variants of the max-flow problem

- The algorithm of Dinic (different implementations reaps better running times than Edmonds-Karp): e.g. simultanous saturation of augmenting paths of the same length O(|V| |E|) - A second capacity function lim for a lower limit of the flow function f with lim(u,v) < f(u,v) < c(u,v)
- A cost function cost where each edge (u,v) has got a second weight cost(u,v) and the min-cost max-flow problem - Networks with multiple sources and sinks
Networks with multiple sources and sinks

s1
s2
t1 T
supersource
S s3 t2
supersink
s4

Network Flow

Cargado por

Información del documento

Descripción original:

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

Network Flow

Cargado por

Copyright:

Formatos disponibles

IT307

Network flow problems

Network flow problems

analysis, running time,

- information through communication network

Example: oil pipeline

Introduction max-flow problem

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Flow below capacity

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

Example: oil pipeline

3) Flow conservation: For all u

3) Flow conservation: For all u

Note: by skew symmetry

Net flow and value of a flow

Net Flow: positive or negative value of f(u,v)

Value of a Flow f: Def: |f| = v V f(s,v)

The max-flow problem

Formal definition of the max-flow problem:

The Ford-Fulkerson method

This method contains 3 important ideas: 1) residual networks

Ford-Fulkerson pseudo code

2 while there exits an augmenting path p

Ford Fulkerson residual networks

residual network Gf = (V,Ef)

cf(u,v) = c(u,v) f(u,v)

Ford Fulkerson residual networks

residual network Gf = (V,Ef)

cf(u,v) = c(u,v) f(u,v)

Ford Fulkerson augmenting paths

cf(p) = min{cf (u,v): (u,v) is on p}

residual network Gf = (V,Ef)

Ford Fulkerson augmenting paths

cf(p) = min{cf (u,v): (u,v) is on p}

residual network Gf = (V,Ef)

Ford Fulkerson augmenting paths

otherwise residual network Gf = (V,Ef)

Ford Fulkerson augmenting paths

otherwise residual network Gf = (V,Ef)

Our virtual flow fp along the augmenting path p in Gf

Ford Fulkerson augmenting the flow

otherwise residual network Gf = (V,Ef)

Our virtual flow fp along the augmenting path p in Gf

Ford Fulkerson new valid flow

cf(p) = min{cf (u,v): (u,v) is on p}

cf(u,v) = c(u,v) f(u,v)

V, we require f(u,v) < c(u,v)

Ford Fulkerson new valid flow

Ford Fulkerson new valid flow

(f + fp) (u ,v) = v V ,v) + fp (u ,v)) (f(u

=v Vf (u ,v) + v Vfp (u ,v) = 0 + 0 = 0

Ford Fulkerson new valid flow

Ford Fulkerson new valid flow

Ford Fulkerson cuts of flow networks