07NetworkFlowI 2x2

7.
N ETWORK F LOW I
7. N ETWORK F LOW I
max-flow and min-cut problems
Ford-Fulkerson algorithm
max-flow min-cut theorem
capacity-scaling algorithm
shortest augmenting paths
blocking-flow algorithm
unit-capacity simple networks
SECTION 7.1
Lecture slides by Kevin Wayne

Copyright 2005 Pearson-Addison Wesley
http://www.cs.princeton.edu/~wayne/kleinberg-tardos
Last updated on Apr 21, 2013 7:35 PM
Flow network
Minimum cut problem
Abstraction for material flowing through the edges.

Digraph G = (V, E) with source s V and sink t V.
Nonnegative integer capacity c(e) for each e E.
Def. A st-cut (cut) is a partition (A, B) of the vertices with s A and t B.

Def. Its capacity is the sum of the capacities of the edges from A to B.
no parallel edges
no edge enters s
no edge leaves t
cap( A, B) =
c(e)
e out of A
capacity
10
15
15
15
10
10
15
10
15
10
16
3
capacity = 10 + 5 + 15 = 30
Minimum cut problem
Minimum cut problem
cap( A, B) =
c(e)
cap( A, B) =
e out of A
c(e)
e out of A
Min-cut problem. Find a cut of minimum capacity.
10
10
don't count edges

from B to A
10
16
capacity = 10 + 8 + 16 = 34
capacity = 10 + 8 + 10 = 28
Maximum flow problem
Def. An st-flow (flow) f is a function that satisfies:
For each e E :
For each v V {s, t} :
0 f (e) c(e)
f (e) = f (e)
e in to v
For each e E :
[capacity]
[flow conservation]
e out of v
0 f (e) c(e)
f (e) = f (e)
e in to v
[capacity]
[flow conservation]
e out of v
f ) f=) = f (e) .
Def. The value of aflow f is:v(val(
e out of s
capacity
flow
inflow at v = 5 + 5 + 0 = 10
outflow at v = 10 + 0
5/9
10
10
0/4
15
/1
5/8
5/5
10
0/4
/6
0 / 15
0 / 15
= 10
5/9
10
10
10 / 10
10
value = 5 + 10 + 10 = 25
10 / 16
7
10
0/4
15
/1
0 / 15
5/8
5/5
10
10
0/4
/6
10
10 / 10
0 / 15
10
10
10 / 16
8

For each e E :
0 f (e) c(e)
f (e) = f (e)
e in to v
7. N ETWORK F LOW I
[capacity]
[flow conservation]
e out of v
f ) f=) = f (e) .
Def. The value of aflow f is:v(val(
e out of s
Max-flow problem. Find a flow of maximum value.
8/9
10
10
0/4
15
0/4
value = 8 + 10 + 10 = 28
/6
0 / 15
8/8
5/5
13
/1
10
0 / 15
10
SECTION 7.1
10 / 10
10
13 / 16
9
Towards a max-flow algorithm
Greedy algorithm.
Greedy algorithm.
Start with f (e) = 0 for all edge e E.

Find an st path P where each edge has f (e) < c(e).
Augment flow along path P.
Repeat until you get stuck.
flow
capacity
0/4
network G

10
0 / 10
0/2
/8
0/9
0/4
network G
0/6
10
0 / 10
value of flow
11
10
0 / 10
0/2
8
/8
0/9
0/6
10
0 / 10
0 +8=8
12
Greedy algorithm.
Greedy algorithm.


0/4
network G
10 / 1
8
2
0/2
/8
0/6
10
8 / 10
0/9
0 / 10
0/4
network G
10
8 + 2 = 10
10
2/2
0 / 10
/8
6
0/6
2/9
6
0
/
10
10 / 10
10 + 6 = 16
13
14
Greedy algorithm.
Greedy algorithm.


ending flow value = 16
0/4
network G
10
but max-flow value = 19
10
6 / 10
2/2
/8
8/9
3/4
network G
6/6
10
10 / 10
10
16
15
10
9 / 10
0/2
/8
9/9
6/6
10
10 / 10
19
16
Residual graph
Original edge: e = (u, v) E.
Capacity c(e).
Flow f (e).
Augmenting path
Def. An augmenting path is a simple st path P in the residual graph Gf .
original graph G
6 / 17
u
flow
Def. The bottleneck capacity of an augmenting P is the minimum

residual capacity of any edge in P.
capacity
Residual edge.
"Undo" flow sent.

e = (u, v) and eR = (v, u).
Residual capacity:
Key property. Let f be a flow and let P be an augmenting path in Gf .
c(e) f (e) if e E
c f (e) =
if e R E
f (e)
Then f ' is a flow and val( f ' ) = val( f ) + bottleneck(Gf, P).
residual
capacity
residual graph Gf
11
AUGMENT (f, c, P)
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
b bottleneck capacity of path P.

FOREACH edge e P
Residual graph: Gf = (V, Ef ).
IF (e E ) f (e) f (e) + b.
Residual edges with positive residual capacity. where flow on a reverse edge
negates flow on a forward edge
Ef = {e : f (e) < c(e)} {eR : f (e) > 0}.
Key property: f ' is a flow in Gf iff f + f ' is a flow in G.
f (eR) f (eR) b.
ELSE
RETURN f.
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
17
18
Ford-Fulkerson algorithm demo
Ford-Fulkerson augmenting path algorithm.
flow
network G

Find an augmenting path P in the residual graph Gf .
capacity
0/4
FORD-FULKERSON (G, s, t, c)
10
0/2
0 / 10
/8
0/6
10
0 / 10
0/9
value of flow
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
FOREACH edge e E : f (e) 0.

Gf residual graph.
residual graph Gf
WHILE (there exists an augmenting path P in Gf )
f AUGMENT (f, c, P).
residual capacity
Update Gf.
RETURN f.
10
10
}
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
s
19
10
10
t
20
network G
network G
0/4
10
0/2
0 / 10
0/4
8
/8
0/6
0 / 10
0/9
10 / 1
8
10
0 +8=8
residual graph Gf
2
0/2
/8
0/6
10
8 + 2 = 10
residual graph Gf
10
10
10
10
network G
network G
10
2/2
0 / 10
/8
22
0/4
0/4
10
10
10
21
8 / 10
0/9
0 / 10
6
0/6
2/9
6
/
10
10 / 10
10
10 + 6 = 16
residual graph Gf
10
0
2/2
6 / 10
/8
6/6
8
/
10
10 / 10
8/9
16 + 2 = 18
residual graph Gf
4
4
6
10
10
7
2
10
10
10
t
23
10
t
24
network G
network G
3
2/4
10
10
0/2
8 / 10
3/4
7
/8
6/6
8/9
9
/
10
min cut
10 / 10
18 + 1 = 19
residual graph Gf
10
10
0/2
9 / 10
6/6
/8
10
10 / 10
9/9
19
1
8
residual graph Gf
10
max flow
nodes reachable from s
10
10
10
25
26
Relationship between flows and cuts
7. N ETWORK F LOW I
Flow value lemma. Let f be any flow and let (A, B) be any cut. Then, the net
flow across (A, B) equals the value of f.
f (e) f (e) = v( f )
e out of A
e in to A
net flow across cut = 5 + 10 + 10 = 25
5/9
SECTION 7.2
10

s
10
0/4
15
/1
0 / 15
5/8
5/5
10
0/4
/6
10
10 / 10
0 / 15
10
value of flow = 25
10
10 / 16
28
f (e) f (e) = v( f )
e out of A
f (e) f (e) = v( f )
e in to A
e out of A
net flow across cut = 10 + 5 + 10 = 25
e in to A
net flow across cut = (10 + 10 + 5 + 10 + 0 + 0) (5 + 5 + 0 + 0) = 25
5/9
5/9
edges from B to A
10
10
0/4
15
0 / 15
5/8
5/5
10
/1
0/4
10
10
10 / 10
0 / 15
/6
10
value of flow = 25
10
0/4
10
0 / 15
5/8
5/5
10
/1
0/4
15
10 / 16
10
10 / 10
0 / 15
/6
10
value of flow = 25
10
10 / 16
29
30
Weak duality. Let f be any flow and (A, B) be any cut. Then, v( f ) cap(A, B).

Pf.
f (e) f (e) = v( f )
e out of A
Pf.
e in to A
v( f ) =
v( f ) =
by flow conservation, all terms

except v = s are 0
f (e)
v(
f) =
v(
=
v(
v( fff ))) =
=
flow-value
lemma
(e)
ff (e)
(e)
of AAff (e)
e out
ee out
out of
of A
A
e out
of
c(e)
c(e)
c(e)
of Ac(e)
e out
ee out
of A
e out
out of
of A
A
=
cap(A,
B)
=
cap(A,
=
B)
= cap(A,
cap(A, B)
B)
of ssf (e)
ee out
out of
e out of s
=
=
f (e) f (e)
f
(e)
f
(e)
vv
ee in
A
A ee out
out of
of vv
in to
to vv
v A e out of v
e in to v
=
=
f (e) f (e).
of AAf (e) ee in
to AAf (e).
ee out
out of
in to
e out of A
(e)

(e)
ff (e)
ff (e)
(e)
ee in
(e)
of
to
e out
Aff (e)
Aff (e)
out of
of A
in to
to A
eee out
ee in
out of A
A
in to A
A
8/9
10
e in to A
10
0/4
15
/1
0 / 15
0/4
/6
10
9 / 10
7/8
5/5
12
0 / 15
10
10
s
10
15
12 / 16
31
value of flow = 27
capacity of cut = 30
32
Max-flow min-cut theorem
Augmenting path theorem. A flow f is a max-flow iff no augmenting paths.
Augmenting path theorem. A flow f is a max-flow iff no augmenting paths.
Max-flow min-cut theorem. Value of the max-flow = capacity of min-cut.
Max-flow min-cut theorem. Value of the max-flow = capacity of min-cut.
Pf. The following three conditions are equivalent for any flow f :
Pf. The following three conditions are equivalent for any flow f :
i. There exists a cut (A, B) such that cap(A, B) = val(f ).
i. There exists a cut (A, B) such that cap(A, B) = val(f ).
ii. f is a max-flow.
ii. f is a max-flow.
iii. There is no augmenting path with respect to f.
iii. There is no augmenting path with respect to f.
[i
[ ii
ii ]
~ii.
Suppose that there is an augmenting path with respect to f.

Can improve flow f by sending flow along this path.
Thus, f is not a max-flow.
Suppose that (A, B) is a cut such that cap(A, B) = val(f ).

Then, for any flow f ', val(f ') cap(A, B) = val(f ).
Thus, f is a max-flow.
weak duality
iii ] We prove contrapositive: ~iii
by assumption
33
34

[ iii
i]
7. N ETWORK F LOW I
Let f be a flow with no augmenting paths.

Let A be set of nodes reachable from s in residual graph Gf.
By definition of cut A, s A.
By definition of flow f, t A.
edge e = (v, w) with v B, w A
original network G
v( f ) =
f (e) f (e)
v( f ) = e outof Af (e) e into Af (e)
v( f ) = e outof Af (e) e into Af (e)
= e outof Ac(e) e in to A
flow-value
= e outof Ac(e)
lemma
= e outof Ac(e)
= cap(A,
e out of A B)
= cap(A, B)
= cap(A, B)

must have f(e) = 0

s
SECTION 7.3
edge e = (v, w) with v A, w B

must have f(e) = c(e)
35
Running time
Bad case for Ford-Fulkerson
Assumption. Capacities are integers between 1 and C.
Q. Is generic Ford-Fulkerson algorithm poly-time in input size?

m, n, and log C
Integrality invariant. Throughout the algorithm, the flow values f (e)
A. No. If max capacity is C, then algorithm can take C iterations.
svwt
swvt
svwt
swvt
svwt
swvt
and the residual capacities cf (e) are integers.

Theorem. The algorithm terminates in at most val (f *) n C iterations.
Pf. Each augmentation increases the value by at least 1.
Corollary. The running time of Ford-Fulkerson is O(m n C).
Corollary. If C = 1, the running time of Ford-Fulkerson is O(m n).
each augmenting path

sends only 1 unit of flow
(# augmenting paths = 2C)
Integrality theorem. Then exists a max-flow f * for which every
flow value f *(e) is an integer.

Pf. Since algorithm terminates, theorem follows from invariant.
37
38
Choosing good augmenting paths
Choosing good augmenting paths
Use care when selecting augmenting paths.
Choose augmenting paths with:
Some choices lead to exponential algorithms.

Clever choices lead to polynomial algorithms.
If capacities are irrational, algorithm not guaranteed to terminate!
Max bottleneck capacity.

Sufficiently large bottleneck capacity.
Fewest number of edges.
Goal. Choose augmenting paths so that:
Theoretical Improvements in Algorithmic Efficiency

for Network Flow Problems
Can find augmenting paths efficiently.

Few iterations.
JACK EDMONDS
University of Waterloo, Waterloo, Ontario, Canada

AND
RICHARD
M. K A R P
University of California, Berkeley, California

ABSTRACT. This paper presents new algorithms for t h e m a x i m u m flow problem, the Hitchcock
t r a n s p o r t a t i o n problem, and t h e general m i n i m u m - c o s t flow problem. U p p e r bounds on the
numbers of steps in these algorithms are derived, and are shown to compale favorably with
upper bounds on t h e numbers of steps required by earlier algorithms.
First, the paper states the m a x i m u m flow problem, gives the F o r d - F u l k e r s o n labeling method
for its solution, and points out t h a t an improper choice of flow a u g m e n t i n g p a t h s can lead to
severe c o m p u t a t i o n a l difficulties. T h e n rules of choice t h a t avoid these difficulties are given.
We show t h a t , if each flow a u g m e n t a t i o n is made along an a u g m e n t i n g p a t h h a v i n g a minimum
n u m b e r of arcs, t h e n a m a x i m u m flow in an n-node network will be o b t a i n e d a f t e r no more t h a n
~(n a - n) a u g m e n t a t i o n s ; and t h e n we show t h a t if each flow change is chosen to produce a
m a x i m u m increase in the flow value then, provided the capacities are integral, a m a x i m u m flow
will be d e t e r m i n e d within at most 1 + logM/(M--1) if(t, S) a u g m e n t a t i o n s , wheref*(t, s) is the
value of the maximum flow and M is the m a x i m u m n u m b e r of arcs across a cut.
Next a new algorithm is given for the m i n i m u m - c o s t flow problem, in which all s h o r t e s t - p a t h
c o m p u t a t i o n s are performed on networks with all weights nonnegative. In particular, this
a l g o r i t h m solves the n X n assigmnent problem in O(n3) steps. Following t h a t we explore a
" s c a l i n g " technique for solving a minimum-cost flow problem by t r e a t i n g a sequence of derived
problems w i t h "scaled d o w n " capacities. It is shown t h a t , using this technique, the solution of
a I i i t c h c o c k t r a n s p o r t a t i o n problem w i t h m sources and n sinks, m ~ n, and m a x i m u m flow B,
requires at most (n + 2) log2 (B/n) flow a u g m e n t a t i o n s . Similar results are also given for the
general minimum-cost flow problem.
An a b s t r a c t s t a t i n g the main results of the present paper was presented at the Calgary
I n t e r n a t i o n a l Conference on C o m b i n a t o r i a l Structures and T h e i r Applications, J u n e 1969.
In a paper b y l)inic (1970) a result closely related to the main result of Section 1.2 is obtained.
Edmonds-Karp 1972 (USA)
39
Dinic 1970 (Soviet Union)
40
Capacity-scaling algorithm
Capacity-scaling algorithm
Intuition. Choose augmenting path with highest bottleneck capacity:

it increases flow by max possible amount in given iteration.
CAPACITY-SCALING(G, s, t, c)
Don't worry about finding exact highest bottleneck path.

Maintain scaling parameter .
Let Gf () be the subgraph of the residual graph consisting only of
__________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
FOREACH edge e E : f (e) 0.

largest power of 2 C.
arcs with capacity .
WHILE ( 1)
Gf () -residual graph.
0
11
11
10
10
WHILE (there exists an augmenting path P in Gf ())

Update Gf ().
/ 2.
0
RETURN f.
17
17
12
12
__________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
Gf
Gf (), = 100
41
42
Capacity-scaling algorithm: proof of correctness
Capacity-scaling algorithm: analysis of running time
Assumption. All edge capacities are integers between 1 and C.
Lemma 1. The outer while loop repeats 1 + log2 C times.

Pf. Initially C / 2 < C; decreases by a factor of 2 in each iteration.
Integrality invariant. All flow and residual capacity values are integral.
Lemma 2. Let f be the flow at the end of a -scaling phase. Then,
Theorem. If capacity-scaling algorithm terminates, then f is a max-flow.
the value of the max-flow val( f ) + m .
proof on next slide
Pf.
By integrality invariant, when = 1 Gf () = Gf .

Upon termination of = 1 phase, there are no augmenting paths.
Lemma 3. There are at most 2m augmentations per scaling phase.
Pf.
Let f be the flow at the end of the previous scaling phase.

LEMMA 2 val( f *) val( f ) + 2 m .
Each augmentation in a -phase increases val( f ) by at least .
Theorem. The scaling max-flow algorithm finds a max flow in O(m log C)
augmentations. It can be implemented to run in O(m2 log C) time.
Pf. Follows from LEMMA 1 and LEMMA 3.
43
44
Capacity-scaling algorithm: analysis of running time

Lemma 2. Let f be the flow at the end of a -scaling phase. Then,
7. N ETWORK F LOW I
the value of the max-flow val( f ) + m .

Pf.
We show there exists a cut (A, B) such that cap(A, B) val( f ) + m .

Choose A to be the set of nodes reachable from s in Gf ().
By definition of cut A, s A.
edge e = (v, w) with v B, w A
By definition of flow f, t A.
must have f(e)

original network
val(
v( ff )) =
v( f )
v( f )
v( f )
=
=
=
=
f (e) f (e)
e out
of A f (e) e in
to A f (e)
e out
of A f (e) e in
to A f (e)
(c(e)
e out
of A f (e) e in
to A f (e)
e out
of A(c(e) ) e in
to A
e out of A
e in to A
e out
of A(c(e) ) e in
to A
c(e)
e out
of A(c(e) ) e in
to A
e out
of Ac(e) e out
of A e in
to A
e out of A
e in to A
e out
of Ac(e) e out
of A e into A
cap(A,
B)
m
of A e into A
e out
of Ac(e) e out
cap(A,
B) - em
e out of A
out of A
e in to A
cap(A, B) - m
cap(A, B) - m
SECTION 17.2
edge e = (v, w) with v A, w B

must have f(e) c(e)
45
Shortest augmenting path
Shortest augmenting path: overview of analysis
Q. Which augmenting path?
L1. Throughout the algorithm, length of the shortest path never decreases.
A. The one with the fewest number of edges.

L2. After at most m shortest path augmentations, the length of the shortest
can find via BFS
augmenting path strictly increases.

Theorem. The shortest augmenting path algorithm runs in O(m2 n) time.
SHORTEST-AUGMENTING-PATH(G, s, t, c)
Pf.
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
O(m + n) time to find shortest augmenting path via BFS.

O(m) augmentations for paths of length k.
If there is an augmenting path, there is a simple one.
FOREACH e E : f (e) 0.
Gf residual graph.
WHILE (there exists an augmenting path in Gf )
P BREADTH-FIRST-SEARCH (Gf, s, t).
1 k < n
O(m n) augmentations.
Update Gf.
RETURN f.
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
47
48
Shortest augmenting path: analysis
Def. Given a digraph G = (V, E) with source s, its level graph is defined by:
Def. Given a digraph G = (V, E) with source s, its level graph is defined by:
(v) = number of edges in shortest path from s to v.
(v) = number of edges in shortest path from s to v.
LG = (V, EG) is the subgraph of G that contains only those edges (v,w) E
LG = (V, EG) is the subgraph of G that contains only those edges (v,w) E
with (w) = (v) + 1.
with (w) = (v) + 1.

Property. Can compute level graph in O(m + n) time.
graph G
Pf. Run BFS; delete back and side edges.

Key property. P is a shortest sv path in G iff P is an sv path LG.
s
level graph LG
level graph LG
= 0
= 1
= 2
= 3
s
49
= 0
= 1
= 2
= 3
50
Let f and f ' be flow before and after a shortest path augmentation.
Let L and L' be level graphs of Gf and Gf ' .
Only back edges added to Gf '
The bottleneck edge(s) is deleted from L after each augmentation.

No new edge added to L until length of shortest path strictly increases.
(any path with a back edge is longer than previous length)

level graph L
level graph L
= 0
= 1
= 2
= 0
= 3
level graph L'
= 1
= 2
= 3
level graph L'
51
52
Shortest augmenting path: review of analysis
Shortest augmenting path: improving the running time
Note. (m n) augmentations necessary on some networks.
Try to decrease time per augmentation instead.

Simple idea O(m n2 )

[Dinic 1970]
Dynamic trees O(m n log n) [Sleator-Tarjan 1983]
JOURNAL
Theorem. The shortest augmenting path algorithm runs in O(m2 n) time.
OF COMPUTER
Pf.
AND
SYSTEM
26, 362-391
SCIENCES
A Data Structure
O(m + n) time to find shortest augmenting path via BFS.

O(m) augmentations for paths of exactly k edges.
O(m n) augmentations.
DANIEL
(1983)
for Dynamic Trees
D. SLEATOR AND ROBERT ENDRE TARJAN
Bell Laboratories,
Received
Murray
May
8, 1982;
Hill,
revised
New Jersey
October
07974
18, 1982
A data structure
is proposed
to maintain
a collection
of vertex-disjoint
trees under a
sequence
of two kinds of operations:
a link operation
that combines
two trees into one by
adding an edge, and a cut operation
that divides one tree into two by deleting an edge. Each
operation
requires
O(log n) time. Using this data structure,
new fast algorithms
are obtained
for the following
problems:
(1)
flows,
Computing
(2)
Solving
and acyclic
nearest
various
flows.
(3)
Computing
(4)
Implementing
The most
maximum
algorithm
common
network
certain
kinds
the network
ancestors.
flow
problems
of constrained
simplex
including
minimum
algorithm
finding
maximum
spanning
for minimum-cost
flows,
blocking
trees.
flows.
significant
application
is (2); an O(mn log n)-time
algorithm
is obtained
to find a
flow in a network
of n vertices and m edges, beating by a factor of log n the fastest
previously
known for sparse graphs.
1. INTR~DIJCTI~N
53
54
In this paper we consider the following problem: We are given a collection of

vertex-disjoint rooted trees. We want to represent the trees by a data structure that
allows us to easily extract certain information about the trees and to easily update the
structure to reflect changes in the trees caused by three kinds of operations:
link(v, w):
If u is a tree root and w is a vertex in another tree, link the trees
containing v and w by adding the edge(v, w), making w the parent of v.

Blocking-flow algorithm
cut(v): If node v is not a tree root, divide the tree containing v into two trees by
deleting the edge from v to its parent.
ever-t(v):
the tree.
7. N ETWORK F LOW I
Turn the tree containing vertex u inside out by making v the root of
Two types of augmentations.

We propose a data structure
that solves this dynamic trees problem. We give two

versions of the data structure. The first has a time bound of O(log n) per operation
when the time is amortized over a worst-case sequence of operations; the second,
362
Normal: length of shortest path does not change.

Special: length of shortest path strictly increases.
0022-0000/83
$3.00
Copyright 0 1983 by Academic Press, Inc.

All rights of reproduction in any form reserved.

Phase of normal augmentations.
Explicitly maintain level graph LG.

Start at s, advance along an edge in LG until reach t or get stuck.
If reach t, augment and and update LG.
If get stuck, delete node from LG and go to previous node.
SECTION 18.1
t
level graph LG
56
Special:


length of shortest path strictly increases.


advance
augment
level graph LG
t
level graph LG
57
58




advance
retreat
level graph LG
t
level graph LG
59
60
Special:




advance
augment
level graph LG
t
level graph LG
61
62




advance
retreat
level graph LG
t
level graph LG
63
64
Special:




retreat
end of phase
level graph LG
t
level graph LG
65
66
Blocking-flow algorithm: analysis

Lemma. A phase can be implemented in O(m n) time.
INITIALIZE(G, s, t, f, c)
ADVANCE(v)
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
LG level-graph of Gf.
IF
Initialization happens once per phase.

At most m augmentations per phase.
(v = t)
P .
AUGMENT(P).
GOTO ADVANCE(s).
Remove saturated edges from LG.
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
Pf.
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
O(mn) per phase
(because an augmentation deletes at least one edge from LG)
At most n retreats per phase.
P .
O(m + n) per phase
(because a retreat deletes one node from LG)
GOTO ADVANCE(s).
RETREAT(v)
O(m) using BFS
At most m n advances per phase.
O(mn) per phase
(because at most n advances before retreat or augmentation)
IF (there exists edge (v, w) LG)
IF (v = s) STOP.
Add edge (v, w) to P.
Theorem. [Dinic 1970] The blocking-flow algorithm runs in O(mn2) time.
ELSE
GOTO ADVANCE(w).
Pf.
Delete v (and all incident edges) from LG.

Remove last edge (u, v) from P.
ELSE GOTO RETREAT(v).
GOTO ADVANCE(u).
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
By lemma, O(mn) time per phase.

At most n phases (as in shortest augment path analysis).
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
67
68
Choosing good augmenting paths: summary
Maximum flow algorithms: theory
Assumption. Integer capacities between 1 and C.
method
# augmentations
running time
augmenting path
nC
O(m n C)
fattest augmenting path
m log (mC)
O(m2 log n log (mC))
year
method
worst case
discovered by
1951
simplex
O(m3 C)
Dantzig
1955
augmenting path
1970
shortest augmenting path
O(m3)
Dinic, Edmonds-Karp
1970
fattest augmenting path
O(m2 log m log( m C ))
Dinic, Edmonds-Karp
1977
blocking flow
O(m 5/2)
Cherkasky
1978
blocking flow
O(m 7/3)
Galil
O(m2
Ford-Fulkerson
C)
capacity scaling
m log C
O(m2
improved capacity scaling
m log C
O(m n log C)
1983
dynamic trees
shortest augmenting path
mn
O(m2 n)
1985
capacity scaling
O(m2 log C)
Gabow
improved shortest augmenting path
mn
O(m n2 )
1997
length function
O(m3/2 log m log C)
Goldberg-Rao
2012
compact network
O(m2 / log m)
Orlin
dynamic trees
mn
O(m n log n )
O(m)
log C)
O(m2
Sleator-Tarjan
log m)
max-flow algorithms for sparse digraphs with m edges, integer capacities between 1 and C
69
70
Maximum flow algorithms: practice
Push-relabel algorithm (SECTION 7.4). [Goldberg-Tarjan 1988]
Warning. Worst-case running time is generally not useful for predicting or
Increases flow one edge at a time instead of one augmenting path at a time.
comparing max-flow algorithm performance in practice.

Best in practice. Push-relabel method with gap relabeling: O(m 3/2).
A New Approach to the Maximum-Flow Problem

ANDREW V. GOLDBERG
Massachusetts
Institute of Technology, Cambridge,
Massachusetts
AND
On I m p l e m e n t i n g P u s h - R e l a b e l M e t h o d
for the M a x i m u m Flow P r o b l e m
ROBERT E. TARJAN
Princeton University,
Princeton, New Jersey, and AT&T
Bell Laboratories,
Murray
Hill, New Jersey
1 Central Institute for Economics and Mathematics,

Krasikova St. 32, 117418, Moscow, Russia
Abstract. All previously known efftcient maximum-flow algorithms work by finding augmenting paths,
either one path at a time (as in the original Ford and Fulkerson algorithm) or all shortest-length
augmenting paths at once (using the layered network approach of Dinic). An alternative method based
on the preflow concept of Karzanov is introduced. A preflow is like a flow, except that the total amount
flowing into a vertex is allowed to exceed the total amount flowing out. The method maintains a preflow
in the original network and pushes local flow excess toward the sink along what are estimated to be
shortest paths. The algorithm and its analysis are simple and intuitive, yet the algorithm runs as fast as
any other known method on dense.graphs, achieving an O(n)) time bound on an n-vertex graph. By
incorporating the dynamic tree data structure of Sleator and Tarjan, we obtain a version of the algorithm
running in O(nm log(n/m)) time on an n-vertex, m-edge graph. This is as fast as any known method
for any graph density and faster on graphs of moderate density. The algorithm also admits efticient
distributed and parallel implementations. A parallel implementation running in O(nlog n) time using
n processors and O(m) space is obtained. This time bound matches that of the Shiloach-Vishkin
algorithm, which also uses n processors but requires O(n) space.
Categories and Subject Descriptors: F.2.2 [Analysis of Algorithms and Problem Complexity]: Nonnumerical Algorithms and Problems; G.2.2 [Discrete Mathematics]: Graph Theory-graph algorithms;
network problems
General Terms: Algorithms, Design, Theory, Verification
EUROPEAN
JOURNAL
OF OPERATIONAL
RESEARCH
Boris V. Cherkassky 1 and Andrew V. Goldberg 2

ELSEVIER
European Journal of Operational Research 97 (1997) 509-542
cher@eemi.msk.su
2 Computer Science Department, Stanford University
Stanford, CA 94305, USA
goldberg~cs. stanford, edu
Theory and Methodology
Computational investigations of maximum flow algorithms

Ravindra
A b s t r a c t . We study efficient implementations of the push-relabel method

for the maximum flow problem. The resulting codes are faster than the
previous codes, and much faster on some problem families. The speedup
is due to the combination of heuristics used in our implementations. We
also exhibit a family of problems for which the running time of all known
methods seem to have a roughly quadratic growth rate.
K . A h u j a a, M u r a l i K o d i a l a m
b, A j a y K . M i s h r a c, J a m e s B . O r l i n d,.
Department t~'lndustrial and Management Engineering. Indian Institute of Technology. Kanpur, 208 016, India
b AT& T Bell Laboratories, Holmdel, NJ 07733, USA
c KA'F-ZGraduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260, USA
d Sloun School of Management, Massachusetts Institute of Technology. Cambridge. MA 02139. USA
Received 30 August 1995; accepted 27 June 1996
Abstract
71
Introduction
The rnaximum flow problem is a classical combinatorial problem that comes up

in a wide variety of applications. In this paper we study implementations of the
push-rdabel [13, 17] method for the problem.
The basic methods for the maximum flow problem include the network simplex method of Dantzig [6, 7], the augmenting path method of Ford and F~lker-
The maximum flow algorithm is distinguished by the long line of successive contributions researchers have made in
obtaining algorithms with incrementally better worst-case complexity. Some, but not all, of these theoretical improvements
have produced improvements in practice. The purpose of this paper is to test some of the major algorithmic ideas developed
in the recent years and to assess their utility on the empirical front. However, our study differs from previous studies in
several ways. Whereas previous studies focus primarily on CPU time analysis, our analysis goes further and provides
detailed insight into algorithmic behavior. It not only observes how algorithms behave but also tries to explain why
algorithms behave that way. We have limited our study to the best previous maximum flow algorithms and some of the
recent algorithms that are likely to be efficient in practice. Our study encompasses ten maximum flow algorithms and five
72

Computer vision. Different algorithms work better for some dense
7. N ETWORK F LOW I
problems that arise in applications to computer vision.
p.1
In IEEE Transactions on PAMI, Vol. 26, No. 9, pp. 1124-1137, Sept. 2004
VERMA, BATRA: MAXFLOW REVISITED
An Experimental Comparison of
Min-Cut/Max-Flow Algorithms for
Energy Minimization in Vision
MaxFlow Revisited:
An Empirical Comparison of Maxflow
Algorithms for Dense Vision Problems
Yuri Boykov and Vladimir Kolmogorov
Abstract
Tanmay Verma
After [15, 31, 19, 8, 25, 5] minimum cut/maximum flow algorithms on graphs emerged as
tanmay08054@iiitd.ac.in
an increasingly useful tool for exact or approximate energy minimization in low-level vision.
Dhruv Batra
The combinatorial optimization literature provides many min-cut/max-flow algorithms with
dbatra@ttic.edu
different polynomial time complexity. Their practical efficiency, however, has to date been
IIIT-Delhi
Delhi, India
TTI-Chicago
Chicago, USA
studied mainly outside the scope of computer vision. The goal of this paper is to provide an
experimental comparison of the efficiency of min-cut/max flow algorithms for applications
Abstract
in vision. We compare the running times of several standard algorithms, as well as a
Algorithms for finding the maximum amount of flow possible in a network (or maxflow) play a central role in computer vision problems. We present an empirical comparison of different max-flow algorithms on modern problems. Our problem instances arise
from energy minimization problems in Object Category Segmentation, Image Deconvolution, Super Resolution, Texture Restoration, Character Completion and 3D Segmentation. We compare 14 different implementations and find that the most popularly used
implementation of Kolmogorov [5] is no longer the fastest algorithm available, especially
for dense graphs.
new algorithm that we have recently developed. The algorithms we study include both
Goldberg-Tarjan style push-relabel methods and algorithms based on Ford-Fulkerson
style augmenting paths. We benchmark these algorithms on a number of typical graphs
in the contexts of image restoration, stereo, and segmentation. In many cases our new
algorithm works several times faster than any of the other methods making near real-time
performance possible. An implementation of our max-flow/min-cut algorithm is available
upon request for research purposes.
Index Terms Energy minimization, graph algorithms, minimum cut, maximum
flow, image restoration, segmentation, stereo, multi-camera scene reconstruction.

Yuri Boykov is with the Computer Science Department at the University of Western Ontario, Canada,
yuri@csd.uwo.ca. Vladimir Kolmogorov is with Microsoft Research, Cambridge, England, vnk@microsoft.com.
This work was mainly done while the authors were with Siemens Corp. Research, Princeton, NJ.
Introduction
Over the past two decades, algorithms for finding the maximum amount of flow possible in
a network (or max-flow) have become the workhorses of modern computer vision and machine learning from optimal (or provably-approximate) inference in sophisticated discrete
models [6, 11, 27, 30, 32] to enabling real-time image processing [38, 39].
Perhaps the most prominent role of max-flow is due to the work of Hammer [23] and
Kolmogorov and Zabih [27], who showed that a fairly large class of energy functions sum
of submodular functions on pairs of boolean variables can be efficiently and optimally
minimized via a reduction to max-flow. Max-flow also plays a crucial role in approximate
minimization of energy functions with multi-label variables [4, 6], triplet or higher order
terms [26, 27, 35, 37], global terms [30], and terms encoding label costs [11, 32].
Given the wide applicability, it is important to ask which max-flow algorithm should be
used. There are numerous algorithms for max-flow with different asymptotic complexities
and practical run-time behaviour. For an extensive list, we refer the reader to surveys in the
literature [2, 7]. Broadly speaking, there are three main families of max-flow algorithms:
Bipartite matching
73
Unit-capacity simple networks

Def. A network is a unit-capacity simple network if:
Q. Which max-flow algorithm to use for bipartite matching?
1. Augmenting-Path (AP) variants: algorithms [5, 13, 14, 17, 21] that maintain a valid
flow during the algorithm, i.e. always satisfying the capacity and flow-conservation
constraints.
Every edge capacity is 1.

Every node (other than s or t) has either (i) at most one entering edge
Generic augmenting path: O( m | f * | ) = O(m n).

Capacity scaling: O(m2 log U) = O(m2).
Shortest augmenting path: O(m n2).
2012. The copyright of this document resides with its authors.

It may be distributed unchanged freely in print or electronic forms.
or (ii) at most one leaving edge.
Q. Suggests "more clever" algorithms are not as good as we first thought?
Property. Let G be a simple unit-capacity network and let f be a 0-1 flow,
A. No, just need more clever analysis!
then Gf is a unit-capacity simple network.
Next. We prove that shortest augmenting path algorithm can be
Ex. Bipartite matching.
implemented
inJ. CoMavx.
SIAM
O(m n1/2)
time.
Vol. 4, No. 4, December 1975
NETWORK FLOW AND TESTING GRAPH CONNECTIVITY*

SHIMON
EVEN"
AND
R. ENDRE
TARJAN:I:
Abstract. An algorithm of Dinic for finding the maximum flow in a network is described. It is
then shown that if the vertex capacities are all equal to one, the algorithm requires at most O(IV[ 1/2 IEI)
time, and if the edge capacities are all equal to one, the algorithm requires at most O(I VI 2/3. IEI) time.
Also, these bounds are tight for Dinics algorithm.
These results are used to test the vertex connectivity of a graph in O(IVI 1/z. IEI 2) time and the
edge connectivity in O(I V[ 5/3. IEI) time.
Key words. Dinics algorithm, maximum flow, connectivity, vertex connectivity, edge connectivity
Let
75
76
Shortest augmenting path algorithm.
Special augmentation:

delete all edges in augmenting path from L
If
get
stuck,
delete
node
from
L
and
go
to
previous
node.
G
Normal augmentation: length of shortest path does not change.

Theorem. [Even-Tarjan 1975] In unit-capacity simple networks, the shortest
augmenting path algorithm computes a maximum flow in O(m n1/2) time.

Pf.
L1.
L2.
L3.
advance
Each phase of normal augmentations takes O(m) time.

After at most n1/2 phases, | f | | f *| n1/2.
After at most n1/2 additional augmentations, flow is optimal.
level graph LG
77
78

augment

advance
level graph LG
level graph LG
79
80

If
get
stuck,
delete
node
from
L
and
go
to
previous
node.
G
retreat

If
get
stuck,
delete
node
from
L
and
go
to
previous
node.
G
advance
level graph LG
level graph LG
81
82

augment

end of phase
level graph LG
level graph LG
83
84
Unit-capacity simple networks: analysis
LEMMA 2. After at most n1/2 phases, | f | | f *| n1/2.
After n1/2 phases, length of shortest augmenting path is >

Level graph has more than n1/2 levels.
Let 1 h n1/2 be layer with min number of nodes: | Vh |

n1/2.
n1/2.
LEMMA 1. A phase of normal augmentations takes O(m) time.

Pf.
level graph LG for flow f
O(1) per edge since each edge traversed and deleted at most once.
O(1) per node since each node deleted at most once.
O(m) to create level graph LG.
V0
85

LEMMA 2. After at most n1/2 phases, | f | | f *| n1/2.
After n1/2 phases, length of shortest augmenting path is > n1/2.

Level graph has more than n1/2 levels.
Let 1 h n1/2 be layer with min number of nodes: | Vh | n1/2.
Let A = {v : (v) < h} {v : (v) = h and v has 1 outgoing residual edge}.
capf (A, B) | Vh | n1/2 | f | | f *| n1/2.
residual graph Gf
residual edges
V0
V1
Vh
Vn1/2
87
V1
Vh
Vn1/2
86

07NetworkFlowI 2x2

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

07NetworkFlowI 2x2

Transféré par

Droits d'auteur :

Formats disponibles

7.

max-flow and min-cut problems

max-flow and min-cut problems

max-flow min-cut theorem

max-flow min-cut theorem

shortest augmenting paths

shortest augmenting paths

unit-capacity simple networks

unit-capacity simple networks

Lecture slides by Kevin Wayne

Last updated on Apr 21, 2013 7:35 PM

Minimum cut problem

Abstraction for material flowing through the edges.

Def. A st-cut (cut) is a partition (A, B) of the vertices with s A and t B.

Minimum cut problem

Minimum cut problem

Def. A st-cut (cut) is a partition (A, B) of the vertices with s A and t B.

Def. A st-cut (cut) is a partition (A, B) of the vertices with s A and t B.

Min-cut problem. Find a cut of minimum capacity.

don't count edges

Maximum flow problem

Maximum flow problem

Def. An st-flow (flow) f is a function that satisfies:

Def. An st-flow (flow) f is a function that satisfies:

Maximum flow problem

max-flow and min-cut problems

max-flow min-cut theorem

Max-flow problem. Find a flow of maximum value.

shortest augmenting paths

unit-capacity simple networks

Towards a max-flow algorithm

Towards a max-flow algorithm

Start with f (e) = 0 for all edge e E.

Start with f (e) = 0 for all edge e E.

Towards a max-flow algorithm

Towards a max-flow algorithm

Find an st path P where each edge has f (e) < c(e).

Start with f (e) = 0 for all edge e E.

Start with f (e) = 0 for all edge e E.

Towards a max-flow algorithm

Towards a max-flow algorithm

Start with f (e) = 0 for all edge e E.

Start with f (e) = 0 for all edge e E.

ending flow value = 16

but max-flow value = 19

Def. The bottleneck capacity of an augmenting P is the minimum

"Undo" flow sent.

Key property. Let f be a flow and let P be an augmenting path in Gf .

Then f ' is a flow and val( f ' ) = val( f ) + bottleneck(Gf, P).

b bottleneck capacity of path P.

Residual graph: Gf = (V, Ef ).

Ford-Fulkerson algorithm demo

Ford-Fulkerson augmenting path algorithm.

Start with f (e) = 0 for all edge e E.

FOREACH edge e E : f (e) 0.

WHILE (there exists an augmenting path P in Gf )

f AUGMENT (f, c, P).

Ford-Fulkerson algorithm demo

Ford-Fulkerson algorithm demo

Ford-Fulkerson algorithm demo

Ford-Fulkerson algorithm demo

Ford-Fulkerson algorithm demo

Ford-Fulkerson algorithm demo