P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML

CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1
Matroids and the Greedy Algorithm

Matroids are objects that generalize certain combinatorial aspects of linear

dependence of finite sets of points in a vector space. A graph can be encoded
by means of its 0/1-valued vertex-edge incidence matrix. It turns out that, when
this matrix is viewed over GF(2), each linearly independent set of columns
corresponds to a forest in the underlying graph, and vice versa. Therefore, a
fortiori, matroids generalize aspects of graphs. From this viewpoint, Hassler
Whitney founded the subject of matroid theory in 1935.
In a natural sense, matroids turn out to yield the precise structure for which
the most naı̈ve “greedy” algorithm finds an optimal solution to combinatorial-
optimization problems for all weight functions. Therefore, matroid theory is
a natural starting point for studying combinatorial-optimization methods. Fur-
thermore, matroids have algorithmic value well beyond the study of greedy
algorithms (see, for example, Chapter 3).
In addition to the algorithmic importance of matroids, we also use matroids
as a starting point for exploring the power of polytopes and linear-programming
duality in combinatorial optimization.

1.1 Independence Axioms and Examples of Matroids

A matroid M is a finite set E(M) together with a subset I(M) of 2 E(M) that
satisfies the following properties:

Independence Axioms
I1. ∅ ∈ I(M).
I2. X ⊂ Y ∈ I(M) =⇒ X ∈ I(M).
I3. X ∈ I(M), Y ∈ I(M), |Y | > |X | =⇒ ∃ e ∈ Y \ X such that X + e ∈
I(M).

49

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

50 1 Matroids and the Greedy Algorithm

The set I(M) is called the set of independent sets of M. The set E(M) is
called the ground set of M. Property I3 is called the exchange axiom.
What follows are some examples that we will revisit as we proceed.

Example (Linear matroid). Let A be a matrix over a field F, with columns

indexed by the finite set E(A). Let E(M) := E(A), and let I(M) be the set
of X ⊂ E(M) such that the columns of A X are linearly independent. In this
case, we say that M is the linear matroid of A and that A is a representation
of M over F. It is very easy to see that properties I1 and I2 hold. To see how I3
holds, suppose that X + e ∈ / I(M) for every e ∈ Y \ X . Then the columns of
AY are in c.s.( A X ) (the column space or linear span of A X ). Hence, c.s.(AY ) is
a subset of c.s.(A X ). Therefore, the dimension of c.s.( AY ) is no more than that
of c.s.(A X ). Therefore, we have |Y | ≤ |X |. ♠

Let G be a graph with vertex set V (G) and edge set E(G). We denote the
numbers of connected components of G (counting isolated vertices as compo-
nents) by κ(G). For F ⊂ E(G), let G.F (G restricted to F) denote the graph
with V (G.F) := V (G) and E(G.F) := F. A set of edges F of graph G is a
forest if it contains no cycle.

Lemma (Forest components). Let F be a forest of a graph G. Then |F| =

|V (G)| − κ(G.F).

Proof. By induction of |F|. Clearly true for |F| = 0. For the inductive step, we
just observe that, for e ∈ F, κ(G.(F − e)) = κ(G.F) − 1.

Example (Graphic matroid). Let G be a graph. Let E(M) := E(G), and let
I(M) be the set of forests of G. In this case, we say that M is the graphic
matroid of G. It is easy to see that I1 and I2 hold. To see how I3 holds, suppose
that X and Y are forests such that X + e is not a forest for every e ∈ Y \ X .
Then every edge in Y \ X would have both ends in the same component of G.X .
Hence, κ(G.Y ) ≥ κ(G.X ). Therefore, by the Lemma (Forest components), we
have |Y | ≤ |X |. ♠

Example (Uniform matroid). Let E(M) be a finite set, and let r be an integer
satisfying 0 ≤ r ≤ |E(M)|. Let I(M) := {X ⊂ E(M) : |X | ≤ r }. In this case,
we say that M is a uniform matroid. ♠

Example (Direct sum). Let M1 and M2 be matroids with E(M1 ) ∩ E(M2 ) =

∅. Define M by E(M) := E(M1 ) ∪ E(M2 ), and I(M) := {X 1 ∪ X 2 : X 1 ∈
I(M1 ), X 2 ∈ I(M2 )}. Then matroid M is the direct sum of M1 and M2 . ♠

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.2 Circuit Properties 51

A system that respects I1 and I2 but not necessarily I3 is called an inde-

pendence system. As the following example indicates, not every independence
system is a matroid.

Example (Vertex packing on a star). Let G be a simple undirected graph.

Define M by E(M) := V (G), and let I(M) be the set of “vertex packings” of
G – a vertex packing of G is just a set of vertices X with no edges of G between
elements of X . Clearly M is an independence system. To see that M need not
be a matroid consider the n-star graph:

with n ≥ 3. The pair X = {1}, Y = {2, 3, . . . , n} violates I3. ♠

1.2 Circuit Properties

For any independence system, the elements of 2 E(M) \ I(M) are called the
dependent sets of M. We distinguish the dependent sets whose proper subsets
are in I(M). We call these subsets the circuits of M, and we write the set of
circuits of M as

C(M) := {X ⊂ E(M) : X ∈ I(M), X − e ∈ I(M), ∀ e ∈ X }.

For example, if M is the graphic matroid of a graph G, then the circuits of

M are the cycles of G. Single-element circuits of a matroid are called loops; if
M is the graphic matroid of a graph G, then the set of loops of M is precisely
the set of loops of G.

Problem [Graphic =⇒ linear over GF(2)]. Show that if A(G) is the

vertex-edge incidence matrix of G, then the matroid represented by A(G),
with numbers of A(G) interpreted in GF(2), is precisely the graphic matroid
of G.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

52 1 Matroids and the Greedy Algorithm

If M is a matroid, then C(M) obeys the following properties:

Circuit Properties
C1. ∅ ∈ C(M).
C2. X ∈ C(M), Y ∈ C(M), X ⊂ Y =⇒ X = Y .
C3. X ∈ C(M), Y ∈ C(M), X = Y , e ∈ X ∩ Y =⇒ ∃ Z ⊂ (X ∪ Y ) − e
such that Z ∈ C(M).

Properties C1 and C2 follow from I1 and I2 and the definition of C(M).

Theorem (Circuit elimination). If M is a matroid, then C(M) satisfies C3.

Proof. Suppose that X , Y , e satisfy the hypotheses of C3 but that (X ∪ Y ) − e

contains no element of C(M). By C2, Y \ X = ∅, so choose some f ∈ Y \ X .
By the definition of C(M), Y − f ∈ I(M).
Let W be a subset of X ∪ Y that is maximal among all sets in I(M) that
contain Y − f . Clearly f ∈ W . Choose some g ∈ X \ W [the set X \ W is
nonempty because X is a circuit and W ∈ I(M)]. Clearly f and g are dis-
tinct because f ∈ Y \ X . In the following figure W is indicated by the shaded
region.

Y
X

g
e
f

Hence,

|W | ≤ |(X ∪ Y ) \ { f, g}| = |X ∪ Y | − 2 < |(X ∪ Y ) − e|.

Now, applying I3 to W and (X ∪ Y ) − e, we see that there is an element h ∈

((X ∪ Y ) − e) \ W , such that W + h ∈ I(M). This contradicts the maximality
of W .

Problem (Linear circuit elimination). Give a direct proof of C3 for linear

matroids.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.3 Representations 53

Problem (Graphic circuit elimination). Give a direct proof of C3 for

graphic matroids.

Property C3 is called the circuit-elimination property. A system satisfying

properties C1 and C2 but not necessarily C3 is called a clutter.

Example [Vertex packing on a star, continued (see p. 51)]. X := {1, i}

and Y := {1, j} are distinct circuits for 1 = i = j = 1, but {i, j} contains no
circuit. ♠

It should be clear that C(M) completely determines I(M) for any indepen-
dence system. That is, given E(M) and C(M) satisfying C1 and C2, there is
precisely one choice of I(M) that has circuit set C(M) that will satisfy I1 and
I2. That choice is

I(M) := {X ⊂ E(M) : ∃ Y ⊂ X, Y ∈ C(M)}.

Problem (Unique-circuit property). Let M be a matroid. Prove that if

X ∈ I(M) and X + e ∈ I(M), then X + e contains a unique circuit of M.
Give an example to show how this need not hold for a general independence
system.

Problem (Linear unique circuit). Give a direct proof of the unique-circuit

property for linear matroids.

Problem (Graphic unique circuit). Give a direct proof of the unique-

circuit property for graphic matroids.

1.3 Representations
The Fano matroid is the matroid represented over GF(2) by the matrix

1 2 3 4 5 6 7
⎛ ⎞
1 0 0 0 1 1 1
F = ⎝0 1 0 1 0 1 1 ⎠.
0 0 1 1 1 0 1

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

54 1 Matroids and the Greedy Algorithm

Exercise [Linear over GF(2) =⇒ graphic]. Prove that the Fano matroid
is not graphic.

A linear matroid may have many representations. A minimal representation

of M is a representation having linearly independent rows. If A and A are
r × n matrices over the same field, having full row rank, and there is a nonsin-
gular matrix B and a nonsingular diagonal matrix D such that A = B AD,
then A and A are projectively equivalent. It is easy to see that projective
equivalence is an equivalence relation. If A and A are projectively equivalent
then they represent the same matroid; however, the converse is not generally
true.

Proposition (Fano representation). The Fano matroid is representable over

a field if and only if the field has characteristic 2. Moreover, F is the only
minimal representation of the Fano matroid over every characteristic-2 field,
up to projective equivalence.

Proof. If the Fano matroid can be represented over a field F, then it has a
minimal representation over F of the form

1 2 3 4 5 6 7
⎛ ⎞
a11 a12 a13 a14 a15 a16 a17
A = ⎝ a21 a22 a23 a24 a25 a26 a27 ⎠.
a31 a32 a33 a34 a35 a36 a37

The first three columns of A must be linearly independent, so, by using ele-
mentary row operations, we can bring A into the form

1 2 3 4 5 6 7
⎛ ⎞
1 0 0 a14 a15 a16 a17
A = ⎝ 0 1 0 a24 a25 a26 a27 ⎠.
0 0 1 a34 a35 a36 a37

We have a14 = 0, a24 = 0, and a34 = 0, as {2, 3, 4} is a circuit. Similarly, a15 =

0, a25 = 0, a35 = 0, and a16 = 0, a26 = 0, a36 = 0. Finally, a17 = 0, a27 = 0,
and a37 = 0, as {1, 2, 3, 7} is a circuit.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.3 Representations 55

Therefore, any minimal representation of the Fano matroid over a field F, up

to multiplication on the left by an invertible matrix, is of the form

1 2 3 4 5 6 7
⎛ ⎞
1 0 0 0 a b c
⎝0 1 0 d 0 e f ⎠,
0 0 1 g h 0 i

with the letters being nonzeros in the field F. We can bring the matrix into the
form

1 2 3 4 5 6 7
⎛ ⎞
1 0 0 0 1 1 1
⎝0 1 0 1 0 q 1 ⎠,
0 0 1 r s 0 1

with the letters being nonzeros, by nonzero row and column scaling (multiply
row 1 by c−1 , row 2 by f −1 , row 3 by i −1 , column 4 by d −1 f , column 5
by a −1 c, column 6 by b−1 c, column 1 by c, column 2 by f , and column 3
by i).
Now, columns 1, 4, and 7 should be dependent; calculating the determinant
and setting it to 0, we get r = 1. Similarly, the required dependence of columns
2, 5, and 7 implies s = 1, and the dependence of columns 3, 6, and 7 implies
q = 1. Therefore, over any field F, F is the only minimal representation of the
Fano matroid, up to projective equivalence.
Finally, columns 4, 5, and 6 should be dependent, so we get 1 + 1 = 0.
Therefore, the field must have characteristic 2.

The non-Fano matroid arises when the GF(2) representation of the Fano
matroid is used but the numbers are considered as rational. The representation
F, viewed over Q, is projectively equivalent to the rational matrix

1 2 3 4 5 6 7
⎛ ⎞
1 0 0 0 1/2 1/2 1/3
F− = ⎝ 0 1 0 1/2 0 1/2 1/3 ⎠.
1 1 1 1 1 1 1

Let F− be the matrix that we obtain by deleting the last row (of all 1’s) of F− .
The linear dependencies among the columns of F− are the same as the affine
dependencies among the columns of the matrix F− . We can plot the columns

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

56 1 Matroids and the Greedy Algorithm

of F− as points in the Euclidean plane and then visualize the independent sets
of the non-Fano matroid as the sets of points that are affinely independent (in
the plane, this means pairs of points that are not coincident and triples of points
that do not lie on a straight line):

6
4
7

3 5 1

Exercise (Nonrepresentable matroids). First prove that the non-Fano ma-

troid is representable over a field if and only if the characteristic of the
field is not 2, and then prove that there are matroids representable over
no field by taking the direct sum of the Fano matroid and the non-Fano
matroid.

1.4 The Greedy Algorithm

Associated with any independence system M is its rank function rM : 2 E(M) →
R, defined by

rM (X ) := max{|Y | : Y ⊂ X, Y ∈ I(M)}.

We call r (E(M)) the rank of M. A set S ⊂ E(M) such that S ∈ I(M) and
M
|S| = rM (E(M)) is a base of M. We write B(M) for the set of bases of M. It
is a simple matter to find a base of the independence system M when M is
a matroid, provided that we can easily recognize when a set is in I(M). We
simply use a “greedy” algorithm:

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.4 The Greedy Algorithm 57

Cardinality Greedy Algorithm

1. S := ∅. U := E(M).
2. While (U = ∅)
i. choose any e ∈ U ; U := U − e;
ii. if S + e ∈ I(M), then S := S + e.

Throughout execution of the algorithm, S ⊂ E(M) and S ∈ I(M). At termi-

nation, |S| = rM (E(M)) (convince yourself of this by using I2 and I3).
The algorithm need not find a base of M, if M is a general independence
system.

Example [Vertex packing on a star, continued (see pp. 51, 53)]. If 1 is chosen
as the first element to put in S, then no other element can be added, but the only
base of M is {2, 3, . . . , n}. ♠

With respect to a matroid M and weight function c, we consider the problem

of finding maximum-weight independent sets Sk of cardinality k for all k satis-
fying 0 ≤ k ≤ rM (E(M)). This is an extension of the problem of determining
the rank of M; in that case, c({e}) = 1, ∀ e ∈ E(M), and we concern ourselves
only with k = rM (E(M)). A greedy algorithm for the present problem is as
follows:

(Weighted) Greedy Algorithm

1. S0 := ∅. k := 1. U := E(M).
2. While (U = ∅)
i. choose sk ∈ U of maximum weight; U := U − sk ;
ii. if Sk−1 + sk ∈ I(M), then Sk := Sk−1 + sk and k := k + 1.

Next we demonstrate that each time an Sk is assigned, Sk is a maximum-

weight independent set of cardinality k.

Theorem (Greedy optimality for matroids). The Greedy Algorithm finds

maximum-weight independent sets of cardinality k for every k satisfying
1 ≤ k ≤ rM (E(M)).

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

58 1 Matroids and the Greedy Algorithm

Proof. The proof is by contradiction. Note that Sk = {s1 , s2 , . . . , sk } for 1 ≤

k ≤ rM (E(M)). Hence, c(s1 ) ≥ c(s2 ) ≥ · · · ≥ c(sk ). Let Tk = {t1k , t2k , . . . , tkk }
be any maximum-weight independent set of cardinality k, with the elements
numbered so that c(t1k ) ≥ c(t2k ) ≥ · · · ≥ c(tkk ). Suppose that c(Tk ) > c(Sk );
then there exists p, 1 ≤ p ≤ k, such that c(t pk ) > c(s p ). Now, consider the
sets
 
t1k , t2k , . . . , t p−1
k
, t pk ,
{s1 , s2 , . . . , s p−1 }.

Property I3 implies that there is some i, 1 ≤ i ≤ p, such that

tik ∈ {s1 , s2 , . . . , s p−1 },

{s1 , s2 , . . . , s p−1 } + tik ∈ I(M).

Now c(tik ) ≥ c(ti+1

k
) ≥ · · · ≥ c(t pk ) > c(s p ); therefore, tik should have been cho-
sen in preference to s p by the Greedy Algorithm.

Exercise (Maximum-weight spanning tree). Use the Greedy Algorithm,

with respect to the graphic matroid of the following edge-weighted graph,
to find a maximum-weight spanning tree.

10
5 2
a
1 −1

c
−8
0
5

−10

e −9
d

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.4 The Greedy Algorithm 59

The Greedy Algorithm can be used to find a maximum-weight independent

set (with no restriction on the cardinality) by stopping once all positive-weight
elements have been considered in Step 2.i.

Problem (Scheduling). Jobs labeled 1, 2, . . . , n are to be processed by a

single machine. All jobs require the same processing time. Each job j has a
deadline d j and a profit c j , which will be earned if the job is completed by
its deadline. The problem is to find the ordering of the jobs that maximizes
total profit. First, prove that if a subset of the jobs can be completed on time,
then they will be completed on time if they are ordered by deadline. Next,
let E(M) := {1, 2, . . . , n}, and
I(M) := {J ⊂ E(M) : the jobs in J are completed on time}.
Prove that M is a matroid by verifying that I1–I3 hold for I(M), and describe
a method for finding an optimal order for processing the jobs.

Exercise (Scheduling). Solve the scheduling problem with the following

data. The machine is available at 12:00 noon, and each job requires one hour
of processing time.

Job j cj dj
1 20 3:00 p.m.
2 15 1:00 p.m.
3 10 2:00 p.m.
4 10 1:00 p.m.
5 6 2:00 p.m.
6 4 5:00 p.m.
7 3 5:00 p.m.
8 2 4:00 p.m.
9 2 2:00 p.m.
10 1 6:00 p.m.

It is natural to wonder whether some class of independence systems, more

general than matroids, might permit the Greedy Algorithm to always find
maximum-weight independent sets of all cardinalities. The following result
ends such speculation.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

60 1 Matroids and the Greedy Algorithm

Theorem (Greedy characterization of matroids). Let M be an independence

system. If the Greedy Algorithm produces maximum-weight independent sets of
all cardinalities for every (nonnegative) weight function, then M is a matroid.

Proof. We must prove that I(M) satisfies I3. The proof is by contradiction.
Choose Y and X so that I3 fails. We assign weights as follows:

⎧
⎨ 1 + , if e ∈ X
c(e) := 1, if e ∈ Y \ X ,
⎩
0, if e ∈ E(M) \ (X ∪ Y )

with  > 0 to be determined. Because Y is in I(M), the Greedy Algorithm

should find a maximum-weight independent set of cardinality |Y |. With just
|X | steps, the Greedy Algorithm chooses all of X , and then it completes X to
an independent set X of cardinality |Y | by using 0-weight elements, for a total
weight of |X |(1 + ). Now we just take care to choose  < 1/|E(M)|, so that
c(X ) < c(Y ). This is a contradiction.

Problem (Swapping Algorithm)

Swapping Algorithm
1. Choose any S ∈ I(M), such that |S| = k.
2. While (∃ S ∈ I(M) with |S | = k, |S S | = 2 and c(S ) > c(S)): Let
S := S .

Prove that if M is a matroid, then the Swapping Algorithm finds a maximum-

weight independent set of cardinality k.

Exercise [Maximum-weight spanning tree, continued (see p. 58)]. Apply

the Swapping Algorithm to calculate a maximum-weight spanning tree for
the edge-weighted graph of the Maximum-weight spanning tree Exercise.

1.5 Rank Properties

Let E be a finite set, and let M be a matroid with E(M) = E. If r := rM , then
r satisfies the following useful properties:

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.5 Rank Properties 61

Rank Properties
R1. 0 ≤ r (X ) ≤ |X |, and integer valued, ∀ X ⊂ E.
R2. X ⊂ Y =⇒ r (X ) ≤ r (Y ), ∀ X, Y ⊂ E.
R3. r (X ∪ Y ) + r (X ∩ Y ) ≤ r (X ) + r (Y ), ∀ X, Y ⊂ E.

Property R3 is called submodularity. The rank function of a general inde-

pendence system M need only satisfy R1 and R2 and the weaker property of
subadditivity: rM (X ∪ Y ) ≤ rM (X ) + rM (Y ).

Example [Vertex packing on a star, continued (see pp. 51, 53, 57)]. For X :=
{1, i} and Y := {1, j}, with i = j, we have rM (X ) = 1, rM (Y ) = 1, rM (X ∪
Y ) = 2, and rM (X ∩ Y ) = 1. ♠

Problem (Cuts). Let G be a graph, let E := V (G), let c be a nonnegative-


weight function on E(G), and define r (X ) := e∈δG (X ) c(e), for X ⊂ E.
Show that r always satisfies R3, but need not satisfy R1 and R2 [even when
c(e) = 1, for all e ∈ E(G)].

Theorem (Submodularity of matroid rank function). If M is a matroid, then

rM satisfies R3.

Proof. Let J be a maximal independent subset of X ∩ Y . Extend J to J X (JY ),

a maximal independent subset of X (Y , respectively). We have rM (X ∩ Y ) =
|J | = |J X ∩ JY |. If we can show that rM (X ∪ Y ) ≤ |J X ∪ JY |, then R3 follows,
because |J X ∪ JY | + |J X ∩ JY | = |J X | + |JY |. Extend J X to a maximal inde-
pendent subset K of X ∪ Y .

Y
X

K \ JX
J
JX \ J
JY\ J

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

62 1 Matroids and the Greedy Algorithm

Suppose that |K | > |J X ∪ JY |. Because J X \ J is contained in both K and

J X ∪ JY , we have |K \ (J X \ J )| > |JY |. Now, by the choice of J X , we have
that K \ (J X \ J ) is an independent subset of Y . This contradicts the choice of
JY .

Our next goal is to show that R1–R3 characterize the rank functions of
matroids. That is, for every E and r satisfying R1–R3, there is a matroid M
with E(M) = E and rM = r . First, we establish a useful lemma.

Lemma (Closure). Suppose that r : 2 E → R satisfies R2 and R3. If X and Y

are arbitrary subsets of E with the property that r (X + e) = r (X ), ∀ e ∈ Y \ X ,
then r (X ∪ Y ) = r (X ).

Proof. The proof is by induction on k = |Y \ X |. For k = 1 there is nothing to

show. For k > 1, choose e ∈ Y \ X .

2r (X ) = r (X ∪ ((Y \ X ) − e)) + r (X + e) (by the inductive hypothesis)

≥ r (X ∪ Y ) + r (X ) (by R3)
≥ 2r (X ) (by R2).

Therefore, equality must hold throughout, and we conclude that r (X ∪ Y ) =

r (X ).

Theorem (Rank characterization of matroids). Let E be a finite set, and

suppose that r : 2 E → R satisfies R1–R3. Then

I(M) := {Y ⊂ E(M) : |Y | = r (Y )}.

defines a matroid M with E(M) := E and rM = r .

Proof. We show that the choice of I(M) in the statement of the theorem satisfies
I1–I3, and then show that r is indeed the rank function of M.
R1 implies that r (∅) = 0; therefore, ∅ ∈ I(M), and I1 holds for I(M).
Now, suppose that X ⊂ Y ∈ I(M). Therefore, r (Y ) = |Y |. R3 implies that

r (X ∪ (Y \ X )) + r (X ∩ (Y \ X )) ≤ r (X ) + r (Y \ X ),

which reduces to

r (Y ) ≤ r (X ) + r (Y \ X ).

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.6 Duality 63

Using the facts that r (Y ) = |Y |, r (Y \ X ) ≤ |Y \ X |, and r (X ) ≤ |X |, we can

conclude that r (X ) = |X |. Therefore, X ∈ I(M), and I2 holds for I(M).
Next, choose arbitrary X, Y ∈ I(M), such that |Y | > |X |. We prove I3 by
contradiction. If I3 fails, then r (X + e) = r (X ) for all e ∈ Y \ X . Applying the
Closure Lemma, we have r (X ∪ Y ) = r (X ). However, r (X ) = |X | and r (X ∪
Y ) ≥ r (Y ) = |Y | implies |Y | ≤ |X |. Therefore, I3 holds for I(M).
We conclude that M is a matroid on E. Because M is a matroid, it has a
well-defined rank function rM which satisfies

rM (Y ) = max{|X | : X ⊂ Y, |X | = r (X )}.

R2 for r implies that

max{|X | : X ⊂ Y, |X | = r (X )} ≤ r (Y ).

Therefore, we need show only that Y contains a set X with |X | = r (X ) = r (Y ).

Let X be a maximal independent subset of Y . Because X + e ∈ I(M), ∀ e ∈
Y \ X , we have r (X + e) = r (X ), ∀ e ∈ Y \ X . By the Closure Lemma, we can
conclude that r (Y ) = r (X ) = |X |, and we are done.

1.6 Duality
Every matroid M has a natural dual M ∗ with E(M ∗ ) := E(M) and

I(M ∗ ) := {X ⊂ E(M) : E(M) \ X contains a base of M}.

Theorem (Matroid duality). The dual of a matroid is a matroid.

Proof. Clearly M ∗ is an independence system. Therefore, it possesses a well-

defined rank function rM ∗ . First we demonstrate that

rM ∗ (X ) = |X | + rM (E(M) \ X ) − rM (E(M)), ∀ X ⊂ E(M ∗ ).

Let Y be a subset of X that is in I(M ∗ ). By the definition of I(M ∗ ), E(M) \ Y

contains a base B of M. If Y is a (setwise) maximal subset of X that is in I(M ∗ ),
then (X \ B) \ Y is empty (otherwise we could append such elements to Y to
get a larger set). Therefore, a maximal such Y is of the form X \ B for some
base B of M. Now, if Y = X \ B is a maximum cardinality such set, then

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

64 1 Matroids and the Greedy Algorithm

|X ∩ B| must be a small as possible, because all bases of M have the same

cardinality.

X
B

Therefore, for any X ⊂ E(M), let B X be a base of M with |B X ∩ X | as small

as possible. By the choice of B X , we have rM ∗ (X ) = |X \ B X |. Moreover, the
choice of B X dictates that |B X \ X | = rM (E(M) \ X ). Therefore, we have

rM ∗ (X ) = |X \ B X |
= |X | + |B X \ X | − |B X |
= |X | + rM (E(M) \ X ) − rM (E(M)).

We leave verification that r := rM ∗ satisfies R1–R3 as a problem.

Problem (Dual rank function). Verify that r := rM ∗ satisfies R1–R3 when

M is a matroid.

It is clear from the specification of I(M ∗ ) that the bases of M ∗ are precisely
the complements of the bases of M. That is, B(M ∗ ) = {E(M) \ B : B ∈
B(M)}. Therefore, another algorithm for finding a maximum-weight base B of
M, with respect to the weight function c, is to use the Greedy Algorithm to
select a minimum-weight base B ∗ of M ∗ , and then let B := E(M) \ B ∗ . The
choice of algorithm may depend on the structure of the matroid. Indeed, for
graphic matroids, there are specialized algorithms that do not appear to extend
to arbitrary matroids.

Problem (Dual of a linear matroid). Prove that if [I, A] is a representation

of a matroid M, then [−A T , I ] is a representation of the dual matroid M ∗ .

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.6 Duality 65

Exercise [Maximum-weight spanning tree, continued (see pp. 58, 60)].

With respect to the edge-weighted graph of the Maximum-weight spanning
tree Exercise, use the Greedy Algorithm to find a minimum-weight base of
the dual of the graphic matroid of the graph.

Exercise [Scheduling, continued (see p. 59)]. With respect to the Schedul-

ing Exercise, use the Greedy Algorithm to find a minimum-weight base of
the associated dual matroid.

Problem (Cocircuits and coloops). Let M be the graphic matroid of a

graph G. Describe the circuits of M ∗ in terms of G. In particular, describe
the loops of M ∗ in terms of G.

A planar embedding of a graph G is a drawing of G in the plane with no edges

crossing. With respect to a planar embedding of G, we construct the planar dual
G ∗ by having a vertex corresponding to each region of the planar embedding of
G and having an edge consisting of each pair of regions that share a common
edge. Note that G ∗ has a vertex corresponding to the outer region of the planar
embedding of G. Evidently G ∗ is planar as well, and it is easily drawn on top
of the planar embedding of G. As each edge of G ∗ naturally crosses an edge of
G in the pair of planar embeddings, it is natural to label each edge of G ∗ with
the label of G corresponding to the edge that it crosses.

Example (Planar dual). Consider the planar graph G:

1
2
6
3 5

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

66 1 Matroids and the Greedy Algorithm

We construct the planar dual as the graph G ∗ , shown in the following figure
with the hollow vertices and dashed edges:

1
2
5
6

Problem (Dual graphic matroids and planar graphs). Let G be a pla-

nar graph. Take any planar embedding of G and form the planar dual G ∗ .
Prove that the graphic matroid of G ∗ is the dual of the graphic matroid
of G.

Problem (Minors of matroids). For a set F ⊂ E(M), define M\F (read M

delete F) by E(M\F) := E(M) \ F, and I(M \ F) := {X ⊂ E(M) \ F :
X ∈ I(M)}. Clearly, M\F is a matroid. Now, define the matroid M/F (read
M contract F) by M/F := (M ∗ \F)∗ .
a. Show that rM/F (X ) = rM (X ∪ F) − rM (F), ∀ X ⊂ E(M) \ F.
b. Choose JF ⊂ F such that JF ∈ I(M) and |JF | = max{|J | : J ⊂ F, J ∈
I(M)}. Show that I(M/F) = {X ⊂ E(M) \ F : X ∪ JF ∈ I(M)}.
c. Describe how to obtain a representation of M\F and of M/F from a
representation of M.

1.7 The Matroid Polytope

The rank function leads to an appealing characterization of the independent sets
of a matroid M in terms of the extreme points of a polytope. Recall that

PI(M) := conv{x(S) : S ∈ I(M)}.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.7 The Matroid Polytope 67

Theorem (Matroid polytope). For any matroid M,


PI(M) = x ∈ R+E(M) : xe ≤ rM (T ), ∀ T ⊂ E(M) .
e∈T

Proof. For every S, T ⊂ E(M), we have


xe (S) = |S ∩ T | .
e∈T

If S ∈ I(M), then |S ∩ T | ≤ rM (T ), as S ∩ T ⊂ T and S ∩ T ∈ I(M). There-


fore, by convexity, we have e∈T x e ≤ rM (T ) for all x ∈ PI(M) , and we
have

PI(M) ⊂ x ∈ R+E(M) : xe ≤ rM (T ), ∀ T ⊂ E(M) .
e∈T

Hence, it suffices to show that every linear-objective function is maximized

over

x ∈ R+E(M) : xe ≤ rM (T ), ∀ T ⊂ E(M)
e∈T

by a point of PI(M) . Without loss of generality, let E(M) = {1, 2, . . . , n}, and let
c(1) ≥ c(2) ≥ · · · ≥ c(n). Let k be the greatest index among the nonnegative
weights. Let Te := {1, 2, . . . , e} for 1 ≤ e ≤ n, and let T0 = ∅. The Greedy
Algorithm for finding a maximum-weight independent set S can be viewed as
determining its characteristic vector x(S) as

rM (Te ) − rM (Te−1 ), if 1 ≤ e ≤ k
xe (S) := .
0, if k < e ≤ n

The point x(S) is a feasible solution of the linear program


max c(e)xe
e∈E(M)

subject to:
(P) 
xe ≤ rM (T ), ∀ T ⊂ E(M);
e∈T

xe ≥ 0, ∀ e ∈ E(M).

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

68 1 Matroids and the Greedy Algorithm

We can check the feasibility of x(S) by using only properties of rM . Non-

negativity follows from R2. Satisfaction of the rank inequalities follows from
R1–R3:

  
xe (S) = rM (Te ) − rM (Te−1 )
e∈T e∈T :
1≤e≤k

 
≤ rM (Te ∩ T ) − rM (Te−1 ∩ T ) (by R3)
e∈T :
1≤e≤k

= rM (Tk ∩ T ) − rM (∅)

≤ rM (T ) − rM (∅) (by R2)

= rM (T ) (by R1).

The dual of P is the linear program


min rM (T ) yT
T ⊂E(M)

subject to:
(D) 
yT ≥ c(e), ∀ e ∈ E(M)
T : e∈T

yT ≥ 0, ∀ T ⊂ E(M).

E(M)
As for P, we can construct a potential solution y ∈ R2 of D, defined by

⎧
⎨ c(e) − c(e + 1), if T = Te with 1 ≤ e < k
yT := c(k), if T = Tk .
⎩
0, otherwise

We need check only that y is feasible to D and that the objective value of x(S)
in P and that of y in D are equal. Then, by the Weak Duality Theorem, x(S) is
optimal in P. Therefore, every linear function is maximized over


x ∈ R+E(M) : xe ≤ rM (T ), ∀ T ⊂ E(M)
e∈T

by a point of PI(M) .

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.7 The Matroid Polytope 69

Clearly, y is nonnegative. For 1 ≤ e < k, we have

 
k
yT = yT
l
T : e∈T l=e

k−1
= c(k) + (c(l) − c(l + 1))
l=e

k−1 
k
= c(k) + c(l) − c(l)
l=e l=e+1

= c(e),
which is certainly ≥ c(e). For e = k, we have

yT = yT = c(k),
k
T : k∈T

which is certainly ≥ c(k). For e > k, we have


yT = 0,
T : e∈T

which is certainly ≥ c(e), because c(e) < 0 for e > k. Therefore, the solution
y is feasible to D. Finally, we have equality of the objective values because
 
k−1
yT rM (T ) = c(k)rM (Tk ) + (c(l) − c(l + 1))rM (Tl )
T ⊂E(M) l=1


k 
k
= c(l)rM (Tl ) − c(l)rM (Tl−1 )
l=1 l=2

k


= c(l) rM (Tl ) − rM (Tl−1 )
l=1

n
= c(l)xl .
l=1

Exercise (Dual solution). With respect to the edge-weighted graph of the

Maximum-weight spanning tree Exercise (see p. 58), calculate the “dual so-
lution” of the previous proof, and use it to verify optimality of the maximum-
weight forest.

Example [Vertex packing on a star, continued (see pp. 51, 53, 57, 61)]. Let
c(1) = 2 and c(2) = c(3) = · · · = c(n) = 1. Following the definition of x(S)
in the previous proof, x1 (S) = 1, x2 (S) = 0, x3 (S) = x4 (S) = · · · = xn (S) = 1,
which picks out the dependent set S = {1, 3, 4, . . . , n} having weight n,

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

70 1 Matroids and the Greedy Algorithm

whereas the maximum-weight independent set is {2, 3, · · · , n}, which has

weight n − 1. ♠

The proof of the characterization of PI(M) for matroids M can be used to

establish a related result.

Theorem (Greedy optimality for polymatroids). Let r be a function on E :=

{1, 2, . . . , n} satisfying R2, R3, and r (∅) = 0. Suppose that c(1) ≥ c(2) ≥ · · · ≥
c(n). Let k be the greatest index among the nonnegative weights. Then the greedy
solution x ∈ R E defined by
r (Te ) − r (Te−1 ), if 1 ≤ e ≤ k
xe :=
0, if k < e ≤ n
for all e ∈ E solves the linear program

max c(e)xe
e∈E

subject to:

xe ≤ r (T ), ∀ T ⊂ E;
e∈T

xe ≥ 0, ∀ e ∈ E.
Furthermore, if k = n and we drop the inequalities xe ≥ 0, ∀ e ∈ E, then we
can omit the hypothesis that r satisfies R2.

For an independence system M, a set T ⊂ E(M) is inseparable if the only

U ⊂ T for which rM (T ) = rM (U ) + rM (T \ U ) are U = T and U = ∅. Rank
inequalities for sets that are not inseparable are redundant because

x j ≤ rM (T )
j∈T

is the sum of

x j ≤ rM (U )
j∈U

and

x j ≤ rM (T \ U ),
j∈T \U

when rM (T ) = rM (U ) + rM (T \ U ).
For an independence system M, a set T ⊂ E(M) is closed if rM (T + e) =
rM (T ) for no e ∈ E(M) \ T . If M is a matroid, then for every T ⊂ E(M) there
is a unique maximal superset cl M (T ) of T , called the closure (or span) of T ,
such that rM (T ) = rM (cl M (T )).

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.7 The Matroid Polytope 71

Rank inequalities for sets that are not closed are redundant because

x j ≤ rM (T )
j∈T

is the sum of

x j ≤ rM (T )
j∈cl M (T )

and
−x j ≤ 0, ∀ j ∈ cl M (T ) \ T.

Theorem (Facets of a matroid polytope). If M is a matroid and { f } ∈ I(M),

∀ f ∈ E(M), then the rank inequalities for nonempty sets that are closed and in-
separable, together with nonnegativity, provide a minimal description of PI(M) .

Proof. Clearly PI(M) is full dimensional because the |E(M)| + 1 points

x(∅) ∪ {x({e}) : e ∈ E(M)}
are affinely independent. Therefore, each facet-describing valid inequality is
unique up to multiplication by a positive scalar.
Each inequality xe ≥ 0 is facet describing because the |E(M)| points
x(∅) ∪ {x({ f }) : f ∈ E(M) − e}
are affinely independent.
Next, suppose that nonempty T is closed and inseparable and consider

F(T ) := PI(M) ∩ x ∈ R E(M) : xe = rM (T ) .
e∈T

Clearly F(T ) is a nontrivial face of PI(M) . We demonstrate that, up to multi-

plication by a positive scalar, the only linear inequality that describes F(T ) is

e∈T x e ≤ rM (T ). By the Unique Description Theorem, this will demonstrate
that F(T ) is a facet of PI(M) .
Let
 
T := S ∈ I(M) : |S ∩ T | = rM (T ) ,

and let
X (T ) := {x(S) : S ∈ T } ⊂ F(T ).
Observe that S ∈ T if and only if

x(S) ∈ PI(M) ∩ x ∈ R E(M) : xe = rM (T ) .
e∈T

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

72 1 Matroids and the Greedy Algorithm


Let e∈E(M) αe xe ≤ β be an arbitrary inequality that describes F(T ). There-

fore, all points x ∈ X (T ) satisfy e∈E(M) αe xe = β.
Let J be a maximal independent subset of T . Clearly J ∈ T , so
 
(∗) αe xe (J ) = αe = β.
e∈E(M) e∈J

Consider f ∈ E(M) \ T . Because T is closed, we have J + f ∈ I(M); hence,

J + f ∈ T and
 
(∗∗) αe xe (J + f ) = αe = β.
e∈E(M) e∈J + f

Subtracting (∗) from (∗∗), we get α f = 0 for f ∈ E(M) \ T .

Next, we demonstrate that αe = α f for all distinct e, f in T . The
following figure may help. Suppose otherwise. Let T1 = {e ∈ T : αe
is maximized over T }. Let T2 = T \ T1 . Let J2 be a maximal independent sub-
set of T2 . Extend J2 to a maximal independent subset J of T . Let J1 = J \ J2 .
Because T is inseparable we have |J1 | < rM (T1 ) [notice that rM (T ) = |J |,
rM (T2 ) = |J2 |]. Therefore, there is some e ∈ T1 \ J1 such that J1 + e ∈ I(M).
It follows that there is some e ∈ J2 such that J := J + e − e is a maximal
independent-subset of T (notice that J + e contains a unique circuit, and that
circuit is contained in J2 + e; so choose e ∈ J2 to be any element of that circuit).
  
Now, J and J are both in T , but e∈J αe > e∈J αe . Hence, e∈T αe xe (J ) >

e∈T αe x e (J ), which is a contradiction.

T1 T2

αmax αe < αmax

J J2
1

e'

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

1.8 Further Study 73


Therefore, every inequality describing F(T ) has the form α e∈T xe ≤ β.
Plugging in x(J ) for some maximal independent subset of T shows that β =
α · rM (T ). Finally, we find that the result follows by noting that (1) α = 0 would
imply F(T ) = PI(M) , and (2) α < 0 yields an inequality that is not valid.

Problem (Base polytope). Let M be a matroid. Suppose that, for every

pair of elements g = h, there is a circuit containing both. Let PB(M) be the
convex hull of the characteristic vectors of bases of M.
a. Give one (nontrivial) linear equation satisfied by all points in PB(M) .
b. Suppose that

αe xe = β
e∈E(M)

is an equation satisfied by all points in PB(M) . Show that αg = αh for

every pair of elements g = h.
c. Show that dim(PB(M) ) = |E(M)| − 1.
d. Give a complete description of PB(M) as the solution set of your equation
from part a and additional linear inequalities.

Problem (Base polytope with a coloop). Let M be a matroid. Suppose that

f is in every base of M. Suppose that, for every other pair of elements g = h
(both different from f ), there is a circuit of M containing g and h.
a. Give two linearly independent equations satisfied by all points in PB(M) .
b. Suppose that

αe xe = β
e∈E(M)

is an equation satisfied by all points in PB(M) . Show that αg = αh for

every pair of elements g = h, both different from f .
c. Show that dim(PB(M) ) = |E(M)| − 2.
d. Give a complete description of PB(M) as the solution set of your equations
from part a and additional linear inequalities.

1.8 Further Study

The theory of matroids is a beautiful and deep area of combinatorial mathemat-
ics. The book by Oxley (1992) is a wonderful resource for learning about this
subject.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010
P1: IML/SPH P2: IML/SPH QC: IML/SPH T1: IML
CB636-FM CB636-Lee CB636-Lee-v2.cls December 11, 2003 16:30 Char Count= 0

74 1 Matroids and the Greedy Algorithm

There are many theoretical and practical studies of the application of greedy
and local-search algorithms to combinatorial-optimization problems. One start-
ing point is the book by Aarts and Lenstra (1997).
Chapter 13 of the book by Ahuja, Magnanti, and Orlin (1993) describes
the details of efficient implementations of algorithms for the minimum-weight
spanning tree problem.

Cambridge Books Published

https://doi.org/10.1017/CBO9780511616655.004 Online ©online
Cambridge University
by Cambridge Press,
University Press 2010