Max Flow – Min Cut Problem
Directed Graph Applications
Shortest Path Problem(Shortest path from one point to another)
Max Flow problems(Maximum material flow through conduit) Liquid flow through pipes Parts through assembly line Current through electrical network, etc
Flow Networks
Directed graph G = (V, E) Only one Source(s) and one Sink(t) Weight on each edge = Capacity of the edge if (u, v) E then Capacity is non-negative,
i.e. c(u, v) ≥ 0 if (u, v) E then Capacity is assumed zero,
i.e. c(u, v) = 0
Flow Networks and Flows
flow(f) in Flow Network(G) is a real-valued function f: V x V → R
f(u, v) is flow from vertex u to vertex v. flow f(u, v) can be positive, negative or zeroConstraints on flow :1. Capacity constraint :
f(u, v) ≤ c(u, v), for all u, v V2. Skew symmetry constraint :
f(u, v) = - f(v, u)
Flow Networks and Flows (Contd.)
3. Flow conservation constraint :Total net flow at vertex must equal 0.
∑j f(i, j) - ∑
k f(k, i) = 0 for all i V – {s, t}
flow in equals flow out
i
j1
j3
j2
k3
k2
k1
jnk4
Maximum Flow
We refer a flow f as maximum if it is feasible and maximize ∑
k f(s, k).
Where f(s, k) is flow out of source s. Problem:
Objective: To find a maximum flow
s t
2
1
(6/10)
(7/8)
(1/1)
(5/6)
(8/10)(flow/capacity)
Multiple source – Multiple sink Network
Convert this problem to single source - single sink.
s3
s4
s5
s2
s1
t3
t2
t1
103
12
5 15
814 20
6
711
13
18
2
Factories
Warehouses
Conversion to flow Network
Add a super – source with infinite weighted edges emanting out to original sources
Add a super – sink with edges of infinite weight from original sinks
s3
s4
s5
s2
s1
t3
t2
t1
103
12
5 15
8
14 20
6
711
13
18
2
Super-source
Factories
Warehouses
Super-sink
st
Another Example : Find a feasible match
Assign persons to tasks such that each task is assigned to a person and each person is assigned to a task in feasible manner.
1
2
3
4
5
6
7
8
persons tasks
Transformation to maximum flow problem
1
2
3
4
5
6
7
8
persons tasks
s
1
11
1
t
1
1
1
1
Ford-Fulkerson Algorithm for Max Flow Key ingredients:-
Residual Networks Augmenting Paths Cut
Limitations :- Flow should be integral or rational On each iteration residual capacity should be
integral
The Residual Network
s
1
2
t
(8/10) (7/8)
(1/1)
(6/10)(5/6)
s
1
2
t
2 1
141
8
5 6
7
u vf (u, v) / c (u, v)
u vc (u, f) – f (u, v)
f (u, v)
Flow Network
Residual Network
(flow / capacity)
Residual capacity r (u ,v)
Augmenting Paths
An augmenting path is a path from s to t in the residual network.
The Residual capacity of augmented path P is P= min {r(i, j): (i, j) P}
Augmentation along P Add Pin each arc along P inflow network Modify residual capacities in residual networkr(u, v) = r(u, v) - Pandr(v, u) = r(v, u) + Pfor u,
v P
s
1
2
t
2 1
141
8
5 6
7s
1
2
t
2
14
8
6 6
8
Cut
An (S,T)-cut in a flow network G = (V,E) is a partition ofvertices V into two disjoint subsets S and T such thats S, t T
e.g., S = { s, 1 } and T = { 2, t }.The capacity of a cut (S,T) is
CAP(S,T) = uSvT c(u,v)
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
The Ford Fulkerson Maximum Flow Method
Beginx := 0;create the residual network G(x);while there is some directed path from s to t in G(x) dobegin
let P be a path from s to t in G(x); := (P);send units of flow along P; update the residual capacities;
end
end {the flow x is now maximum}.
Proof of Correctness of the Algorithm
Assume that all data are integral.Lemma:
At each iteration all residual capacities are integral.Proof:
By assumption it is true at beginning. Assume it is true after the first k-1 augmentations, and consider augmentation k along path P. The residual capacity of P is the smallest residual capacity on P, which is integral. After updating, we modify residual capacities by 0, or , and thus residual capacities stay integral.
Proof of finiteness of Algorithm
Proof: The capacity of each augmenting path is at least 1. The augmentation reduces the residual capacity of some arc (s, j) and does not increase the residual capacity of (s, i) for any i.So, the sum of the residual capacities of arcs out of s keeps decreasing, and is bounded below by 0.Number of augmentations is O(nU), where U is the largest capacity in the network.
Indication of Optimum Flow
There is no augmenting path in the residual network.
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
s
1
2
t
1
13
9
6 7
8
Indication of Optimum Flow (Contd.)
Flow across the network is equal to the capacity of some cut(Max Flow Min Cut Theorem)
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
Weak Duality Theorem for the Max Flow ProblemTheorem:
If f is any feasible flow and if (S,T) is an (s,t)-cut, then the flow | f | from source to sink in the network is at most CAP(S,T).
Proof: We define the flow across the cut (S,T) to bef(S,T) = iSjT f(i, j) - iSjT f(j, i)
S T
Flows across different cuts
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
s
1
2
t
(9/10) (8/8)
(1/1)
(7/10)(6/6)
S T
S/T cut
If S = {s}, then the flow across (S, T) is 9 + 6 = 15
If S = {s,1}, then the flow across (S, T) is 8 + 1 + 6 = 15
If S = {s,2}, then the flow across (S, T) is 9 + 7 – 1 = 15
More on Flows Across Cuts
Claim: Let (S,T) be any s-t cut. Then f(S,T) = | f | = flow into t.
Proof: Add the conservation of flow constraints for each node i S - {s} to the constraint that the flow leaving s is |f|. The resulting equality is f(S,T) = |f|.
j f(i, j) - k f(k, i) = 0 for each i {S} - s
j f(s, j) = | f | S T
More on Flows Across CutsClaim:
The flow across (S,T) is at most the capacity of a cut.Proof:
If i S, and j T, then f(i, j) c(i, j). If i T, and j S, then f(i, j) 0.
f(S,T) = iSjT f(i, j) - iSjT f(j, i)
CAP(S,T) = iSjT c(i, j) - iSjT 0
Max Flow Min Cut Theorem
Theorem: (Optimality conditions for max flows). The following are equivalent.1. A flow x is maximum.2. There is no augmenting path in G(x). 3. There is an s-t cutset (S, T) whose capacity is the flow value
of x.
Corollary: (Max-flow Min-Cut). The maximum flow value is the minimum value of a cut.
Proof of Theorem: 1 2. (not 2 not 1)Suppose that there is an augmenting path in G(x). Then x is not maximum.
Max Flow Min Cut Theorem Contd.
3 1. Let v = Fx(S, T) be the flow from s to t. By assumption, v = CAP(S, T). By weak duality, the maximum flow is at most CAP(S, T). Thus the flow is maximum.
2 3. Suppose there is no augmenting path in G(x).
Claim: Let S be the set of nodes reachable from s in G(x). Let T = N\S. Then there is no arc in G(x) from S to T.
Thus i S and j T f(i, j) = c(i, j)i T and j S f(i, j) = 0.
It follows that Fx(S,T) = iSjT f(i, j) - iSjT f(j, i)
= iSjT c(i, j) - iSjT 0 = CAP(S,T)
There is no arc from S to T in G(x)
Max Flow Min Cut Theorem Contd.
S T
Saturated
Reachable from s
Not Reachable from s
.s .t0
27
Limitation of Algorithm : example 1
s
1
2
t
M M
MM1
M >> 1
• Termination even with integral flow and few nodes can take large no. of steps.• Depends on path selection and capacity of flow network.
Flow Network
Flow
28
After 1st augmentation
s
1
2
t
M-1 M
M-1M1
1
1
Residual Network
29
After 2nd augmentation
s
1
2
t
M-1 M-1
M-1M-11
1
1
1
1
30
After 3rd augmentation
s
1
2
t
M-2 M-1
M-2M-11
2
2
1
1
31
After 2M augmentations
s
1
2
tM
M
M
M
1
32
s 3
4
t
m r
mm
Flow Network 1
2
m m
m
1
1
m >> 1
1/2 < r < 1
Limitation of Algorithm (Non termination for irrational flow)
3333
Flow in Flow Network Residual Network
3
4
r
1
2 1
1
32 tS 1 1 1
After 1st augmentation
343434
Flow in Flow Network Residual Network
r
r3
4
r
1
2 1-r
1-r
3
4
r
1
2
After 2nd augmentation
35353535
Flow in Flow Network Residual Network
1-r
3
4
r
1
2 1
r r
3
r
1
2
After 3rd augmentation
3636363636Flow in Flow Network Residual Network
3
4
2r -11
2 r
1
1-r1 - r
1-r
3
4
1 - r
1
2 1 - r
After 4th augmentation
373737373737Residual NetworkFlow in Flow Network
3
4
2r -1
1
2 1r
1-r
1-r
3
4
1
2
1 - r
1-r1-r
After 5th augmentation
Non termination in irrational flow
38
3
4
1
2
ai
ai+1
3
4
1
2
ai - ai+1 = ai+2
ai+1 - (ai - ai+1) = ai+3
And this goes on …………….
39
Summary and Extensions
1. Augmenting path theorem
2. Ford-Fulkerson Algorithm
3. Duality Theory.
4. Computational Speedups.