CSE 347 Lecture 4

Maximum flow

1
Ship cement from factory to building
20 10

20
s t
Factory 10 20
Building
Input: site
𝒔: source; 𝒕: destination
Graph with directed edges
weights on each edge: capacity
Goal: Ship as much stuff as possible while obeying
capacity constraints.
2
Solution

20 10 20 10

20 20
s t s t
10 20 10 20

3
 Maximum Flow Problem
Graph: (𝑉, 𝐸) directed and weighted
• Unique source and sink nodes à 𝑠, 𝑡
• Each edge has capacity 𝑐(𝑒) [Integer]
A valid flow assignment assigns an integer 𝑓(𝑒) to each edge s.t.
Capacity constraint: 0 ≤ 𝑓 𝑒 ≤ 𝑐 𝑒
Flow conservation: ∑!∈#!" (%) 𝑓 𝑒 = ∑!∈##$% (%) 𝑓(𝑒), ∀𝑣 ∈ 𝑉 −
𝑠, 𝑡
𝐸'( (𝑣): set of incoming edges to 𝑣
𝐸)*+ 𝑣 : set of outgoing edges from 𝑣
Compute: Maximum Flow: Find a valid flow assignment to
Maximize 𝐹 = ∑ 𝑓 𝑒 =∑ 𝑓 𝑒
!∈# (,) !∈# (+)
)*+ '(
4
Additional assumptions
1. 𝑠 has no incoming edges, 𝑡 has no outgoing edges
2. You do not have a cycle of 2 nodes

u v

5
A proposed algorithm:
1. Find a path from 𝑠 to 𝑡
2. Push as much flow along the path as possible
3. Adjust the capacities
4. Repeat until we cannot find a path

6
A proposed algorithm
1. Find a path from 𝑠 to 𝑡
2. Push as much flow along the path
3. Adjust the capacities
4. Repeat until we cannot find a path

20 10

20
s t s t
10 20

s t s t
7
Does it always get us the maximum flow?

20 10

20
s t s t
10 20

8
What now?
Greedy?
Dynamic Programming?

9
Let’s fix it:

Residual Graph: If there is an edge 𝑒 = 𝑢, 𝑣 in 𝐺, we

will add a back edge 𝑒̅ = (𝑣, 𝑢). Capacity of 𝑒̅ = flow
on 𝑒. Call this graph 𝐺! .
Initialize:

20 10

20
s t
10 20

10
Augmenting Paths Algorithm:
Algorithm:
• Find an “augmenting path” 𝑃.
• 𝑃 can contain forward or backward edges!
• Say the smallest residual capacity along the path is 𝑘.
• Push k flow on the path (𝑓 𝑒 = 𝑓 𝑒 + 𝑘 for all
edges on path 𝑃)
• Reduce the capacity of all edges on the path 𝑃 by k
• Increase the capacity of the corresponding
mirror/back edges
Repeat until there are no augmenting paths
11
Execution of Algorithm

20 10

20
s t s t
10 20

s t s t

12
Formalize: Ford-Fulkerson(FF)
Algorithm
1. Initialize the residual graph 𝐺! = 𝐺
2. Find an augmenting path 𝑃 with capacity 𝑘
3. Fix up the residual capacities in 𝐺!
– 𝑐 𝑒 =
– 𝑐 𝑒̃ =
4. Repeat 2 and 3 until no augmenting path can be
found in 𝐺!

13
Proof of Correctness: Valid Flow
Lemma 1: FF finds a valid flow
– Capacity and conservation constraints are not violated
– Capacity constraint: 0 ≤ 𝑓 𝑒 ≤ 𝑐 𝑒
– Flow conservation: ∑!∈#!"(%) 𝑓 𝑒 = ∑!∈##$%(%) 𝑓(𝑒),
∀𝑣 ∈ 𝑉 − 𝑠, 𝑡
Proof: Induct on augmenting paths
– Base Case: 𝑓 𝑒 = 0 on all edges

14
Inductive Case
Lemma 1: FF finds a valid flow
Inductive Hypothesis:

Inductive Step:

15
Inductive Case
Capacity Constraints: Consider an edge 𝑒 in 𝑃
(a) If 𝑒 is a forward edge (in the original graph)

(b) If 𝑒 is a back edge with residual capacity ≥ 𝑘

16
 Inductive Case
Conservation Constraints: Consider a vertex 𝑣 on path 𝑃

(a) Both forward edges

(b) Both back edges

(c) Redirecting flow

(d) Similar to (c)

17
Proof of Correctness: Termination
Lemma 2: FF terminates
Proof:

18
Runtime of Ford-Fulkerson Algorithm
Runtime:
(time to find a path) × (number of augmenting paths)

Food for thought: Is this indeed polynomial time?

19
Proof of Correctness: Optimality
From Lemma 1 and 2, we know that FF returns a feasible
solution, but does it return the maximum flow?
Max-flow Min-cut Theorem!
• Given a graph 𝐺(𝑉, 𝐸), a graph cut is a partition of
vertices into 2 subsets.
– 𝑺: 𝑠 + maybe some other vertices
– 𝑽 − 𝑺: 𝑡 + maybe some other vertices

20
Graph Cut
Graph cut: a partition of vertices into 2 subsets.
𝑺 = 𝑠 + maybe some other vertices
𝑽 − 𝑺 = 𝑻 = 𝑡 + maybe some other vertices
a

s t s t

21
Graph Cut
Graph cut: a partition of vertices into 2 subsets.
𝑺: 𝑠 + maybe some other vertices
𝑽 − 𝑺 = 𝑻: 𝑡 + maybe some other vertices
Define the cut by 𝑺.
Capacity of the cut:

sum of capacity of edges that go from a vertex in 𝑆 to a

vertex in 𝑇.

22
Cut Capacity
Capacity: 𝐶 𝑆 = ∑"∈$,&∈' 𝑐(𝑢, 𝑣)

20 10

20

s t
10 20

23
Max-flow Min-cut Theorem
Lemma: For all valid flows 𝑓, 𝑓 ≤ 𝐶(𝑆) for all cut 𝑆.
Proof:

Min-cut: cut of smallest capacity, 𝑆 ∗ . 𝑓 ≤ 𝐶(𝑆 ∗ )

Lemma: FF produces a flow = 𝐶(𝑆 ∗ )

24
Max-flow Min-cut Theorem
Min-cut: cut of smallest capacity, 𝑆 ∗ . 𝑓 ≤ 𝐶(𝑆 ∗ )
Lemma: FF produces a flow = 𝐶(𝑆 ∗ )
Proof: Let 𝒇B be the flow found by FF. No augmenting
paths in 𝐺! .
𝒔D : all vertices that can be reached from s using edges with
capacities > 0.

25
Max-flow Min-cut Theorem
• All edges going out of this cut is saturated.
• All real edges out of this cut are saturated (residual
capacity = 0).
• No flow is going into the cut 𝑆. Why?

H so FF produces the max-flow!

• 𝐶 𝑆 ∗ = |𝑓|,
26
New Problem: Bipartite Matching
• n classes and n rooms; we want to match classes to
rooms.

27
New Problem: Bipartite Matching
• Bipartite graph 𝐺 = (𝑉, 𝐸) (unweighted and undirected)
– Vertices are either in set L or R
– Edges only go between vertices of different sets
• Matching: A subset of edges 𝑀 ⊆ 𝐸 s.t.
– Each vertex has at least one edge from M incident on it.
• Maximum Matching: matching of the largest size.
• We will reduce the problem to the problem of finding
the maximum flow

28
 Reduction
• Given a bipartite graph 𝐺 = 𝑉, 𝐸 , construct a graph
𝐺 / = 𝑉 / , 𝐸 / s.t.

29
Proof of Correctness
Claim: G’ has a flow of k iff G has a matching of size k
Proof: Two directions:
Direction 1:

Direction 2:

Corollary: 30
Conclusion: Maximum Flow
• Problem input and target
• Ford-Fulkerson Algorithm
– Execution: residual graph
– Runtime
• FF correctness proof
– Max-flow Min-cut Theorem
– Graph Cut definition
– Capacity of cut
• Reduction to Bipartite Matching

31