Lecture 2

Geometry of LPs∗

Last time we saw that, given a (minimizing) linear program in equational form, one of the
following three possibilities is true:

1. The LP is infeasible.

2. The optimal value of the LP is −∞ (i.e., the LP does not have a bounded optimum).

3. A basic feasible solution exists that achieves the optimal value.

2.1 Finding a basic feasible solution

Suppose we have an LP in equational form:

min{ cT x | Ax = b, x ≥ 0 }, where A is an m × n matrix of rank m.

Recall how we might attempt to find a BFS to this LP: Let B ⊆ [n], with |B| = m, such that
AB (the set of columns of A corresponding to the elements of B) is linearly independent.
(Such a set of columns exists because A has full rank.) Let N = [n] \ B be the indices of the
columns of A that are not in B. Since AB is invertible, we can define a vector x ∈ Rn by

xB = A−1
B b,
xN = 0.

By construction, x is a basic solution. If x is also feasible (i.e., if x ≥ 0), then it is a BFS.

Fact 2.1. Every LP in equational form that is feasible has a BFS. (Note that this BFS may
or may not be optimal.)

Proof. Pick some feasible point x̃ ∈ Rn . (In particular, since x̃ is feasible, x̃ ≥ 0.) Let

P = { j | x̃j > 0 }
*
Lecturer: Anupam Gupta. Scribe: Brian Kell.

1
LECTURE 2. GEOMETRY OF LPS 2

be the set of coordinates of x̃ that are nonzero. We consider two cases, depending on whether
the columns of AP are linearly independent.
Case 1. The columns of AP are linearly independent. Then we may extend P to a
basis B, i.e., a subset P ⊆ B ⊆ [n] with |B| = m such that the columns of AB are also
linearly independent. Let N = [n] \ B; then x̃N = 0 (because P ⊆ B). In addition, since
Ax̃ = b, we have AB x̃B = b, so x̃ = A−1 B b. So x̃ is a basic solution; since it is feasible by
assumption, it is a BFS. (Note, by the way, that the equation x̃ = A−1 B b means that x̃ is the
unique solution to Ax = b having xN = 0.)
Case 2. The columns of AP are linearly dependent. Let N = [n] \ P . Then, by the
definition of linear dependence, there exists a nonzero vector w ∈ Rn with wN = 0 such that
AP wP = 0. For any λ ∈ R, the vector x̃ + λw satisfies A(x̃ + λw) = b, because

A(x̃ + λw) = Ax̃ + λAw = b + 0 = b.

Because x̃N = 0 and wN = 0, we have (x̃ + λw)N = 0, so x̃ + λw has no more nonzero entries
than x̃ does. Since x̃P > 0, for sufficiently small  > 0 both x̃ + w and x̃ − w are feasible
(i.e, x̃ ± w ≥ 0). Let η = sup{  > 0 | x̃ ± w ≥ 0 } be the largest such ; then one of
x̃ ± ηw has one more zero coordinate than x̃ does. We can repeat this until we find a feasible
solution with no more than m nonzero coordinates, at which point Case 1 applies and we
have found a BFS.
(Intuitively, for sufficiently small  > 0, one of x̃ ± w is moving toward a nonnegativity
constraint, that is, toward the boundary of the nonnegative orthant. When  becomes just
large enough that the point x̃ ± w reaches the boundary of the nonnegative orthant, we
have made one more coordinate of the point zero.)

2.2 Geometric definitions

Definition 2.2. Given points x, y ∈ Rn , a point z ∈ Rn is a convex combination of x and y
if
z = λx + (1 − λ)y
for some λ ∈ [0, 1].

Definition 2.3. A set X ⊆ Rn is convex if the convex combination of any two points in X
is also in X; that is, for all x, y ∈ X and all λ ∈ [0, 1], the point λx + (1 − λ)y is in X.

Definition 2.4. A function f : Rn → R is convex if for all points x, y ∈ Rn and all λ ∈ [0, 1]
we have

f λx + (1 − λ)y ≤ λf (x) + (1 − λ)f (y).

Fact 2.5. If P ⊆ Rn is a convex set and f : Rn → R is a convex function, then, for any
t ∈ R, the set
Q = { x ∈ P | f (x) ≤ t }
is also convex.
LECTURE 2. GEOMETRY OF LPS 3

Proof. For all x1 , x2 ∈ Q and all λ ∈ [0, 1], we have

f λx1 + (1 − λ)x2 ≤ λf (x1 ) + (1 − λ)f (x2 ) ≤ λt + (1 − λ)t = t,

so λx1 + (1 − λ)x2 ∈ Q.
Fact 2.6. The intersection of two convex sets is convex.
Proof. Let P, Q ⊆ Rn be convex sets, and let x1 , x2 ∈ P ∩ Q. Let λ ∈ [0, 1]. Because
x1 , x2 ∈ P and P is convex, we have λx1 + (1 − λ)x2 ∈ P ; likewise, λx1 + (1 − λ)x2 ∈ Q. So
λx1 + (1 − λ)x2 ∈ P ∩ Q.
Definition 2.7. A set S ⊆ Rn is a subspace if it is closed under addition and scalar multi-
plication.
Equivalently, S is a subspace if S = { x ∈ Rn | Ax = 0 } for some matrix A.
Definition 2.8. The dimension of a subspace S ⊆ Rn , written dim(S), is the size of the
largest linearly independent set of vectors contained in S.
Equivalently, dim(S) = n − rank(A).
Definition 2.9. A set S 0 ⊆ Rn is an affine subspace if S 0 = { x0 + y | y ∈ S } for some
subspace S ⊆ Rn and some vector x0 ∈ Rn . In this case the dimension of S 0 , written dim(S 0 ),
is defined to equal the dimension of S.
Equivalently, S 0 is an affine subspace if S 0 = { x ∈ Rn | Ax = b } for some matrix A and
some vector b.
Definition 2.10. The dimension of a set X ⊆ Rn , written dim(X), is the dimension of the
minimal affine subspace that contains X.
Note that if S10 and S20 are two affine subspaces both containing X, then their intersec-
tion S10 ∩ S20 is an affine subspace containing X. Hence there is a unique minimal affine
subspace that contains X, so dim(X) is well defined.
Equivalently, given x0 ∈ X, the dimension of X is the largest number k for which there
exist points x1 , x2 , . . . , xk ∈ X such that the set {x1 − x0 , x2 − x0 , . . . , xk − x0 } is linearly
independent.
Note that the definition of the dimension of a set X agrees with the definition of the
dimension of an affine subspace if X happens to be an affine subspace, and the definition
of the dimension of an affine subspace S 0 agrees with the definition of the dimension of a
subspace if S 0 happens to be a subspace.
Definition 2.11. A set H ⊆ Rn is a hyperplane if H = { x ∈ Rn | aT x = b } for some
nonzero a ∈ Rn and some b ∈ R.
A hyperplane is an affine subspace of dimension n − 1.
Definition 2.12. A set H 0 ⊆ Rn is a (closed) halfspace if H 0 = { x ∈ Rn | aT ≥ b } for some
nonzero a ∈ Rn and some b ∈ R.
LECTURE 2. GEOMETRY OF LPS 4

A hyperplane can be written as the intersection of two halfspaces:

{ x ∈ Rn | aT x = b } = { x ∈ Rn | aT x ≥ b } ∩ { x ∈ Rn | −aT x ≥ −b }.

Both hyperplanes and halfspaces are convex sets. Therefore the feasible region of an LP is
convex, because it is the intersection of halfspaces and hyperplanes. The dimension of the
feasible region of an LP in equational form, having n variables and m linearly independent
constraints (equalities), is no greater than n − m, because it is contained in the intersection
of m distinct hyperplanes, each of which is an affine subspace of dimension n − 1. (The
dimension of the feasible region may be less than n − m, because of the nonnegativity
constraints, for instance.)
For example, the region in R3 defined by

x1 + x2 + x3 = 1,
x≥0

is a 2-dimensional triangle; here, n − m = 3 − 1 = 2. (Note, however, that if the constraint

were x1 + x2 + x3 = 0, the region would have dimension 0.)

Figure 2.1: The region { x ∈ R3 | x1 + x2 + x3 = 1, x ≥ 0 }.

Definition 2.13. A polyhedron in Rn is the intersection of finitely many halfspaces.

For example, feasible regions of LPs are polyhedra.
Definition 2.14. A polytope is a bounded polyhedron, that is, a polyhedron P for which
there exists B ∈ R+ such that kxk2 ≤ B for all x ∈ P .
LECTURE 2. GEOMETRY OF LPS 5

Both polyhedra and polytopes are convex.

Definition 2.15. Given a polyhedron P ⊆ Rn , a point x ∈ P is a vertex of P if there exists

c ∈ Rn such that cT x < cT y for all y ∈ P , y 6= x.

Suppose x̂ is a vertex of a polyhedron P ⊆ Rn . Let c be as in the definition above.

Take K = cT x̂. Then for all y ∈ P we have cT y ≥ K, so the polyhedron P is contained
in the halfspace { x ∈ Rn | cT x ≥ K }, i.e., P lies entirely on one side of the hyperplane
{ x ∈ Rn | cT x = K }. Furthermore, the vertex x̂ is the unique minimizer of the function cT z
for z ∈ P .

Definition 2.16. Given a polyhedron P ⊆ Rn , a point x ∈ P is an extreme point of P if

there do not exist points u, v 6= x in P such that x is a convex combination of u and v.

In other words, x is an extreme point of P if, for all u, v ∈ P ,

x = λu + (1 − λ)v for some λ ∈ [0, 1] implies u = v = x.

2.3 Equivalence of vertices, extreme points, and basic

feasible solutions
In fact, vertices and extreme points are the same thing, and for an LP the vertices (i.e.,
extreme points) of its feasible region are precisely its basic feasible solutions. This is shown
in the following theorem.

Theorem 2.17. Consider an LP in equational form, i.e., min{ cT x | Ax = b, x ≥ 0 }, and

let K be its feasible region. Then the following are equivalent:

1. The point x is a vertex of K.

2. The point x is an extreme point of K.

3. The point x is a BFS of the LP.

Proof. (1 ) ⇒ (2 ). Let x be a vertex of K. Then there exists c ∈ Rn such that cT x < cT y

for all y ∈ K, y 6= x. Suppose for the sake of contradiction that x is not an extreme point
of K, so there exist some points u, w ∈ K with u, w 6= x and some λ ∈ [0, 1] such that
x = λu + (1 − λ)w. Then

cT x = λcT u + (1 − λ)cT w < λcT x + (1 − λ)cT x = cT x,

which is a contradiction. Hence x is an extreme point of K.

(2 ) ⇒ (3 ). Let x be an extreme point of K. (In particular, x is a feasible solution for
the LP, so x ≥ 0.) Let P = { j | xj > 0 } be the set of nonzero coordinates of x. We consider
two cases, depending on whether AP (the set of columns of A corresponding to P ) is linearly
independent.
LECTURE 2. GEOMETRY OF LPS 6

Case 1. The columns of AP are linearly independent. Then x is a BFS. (This is the same
as in the proof of Fact 2.1: Extend P to a basis B, and let N = [n] \ B; then xB = A−1 B b
and xN = 0.)
Case 2. The columns of AP are linearly dependent. Then there exists a nonzero vector wP
such that AP wP = 0. Let N = [n] \ P and take wN = 0. Then Aw = AP wP + AN wN = 0.
Now consider the points y + = x + λw and y − = x − λw, where λ > 0 is sufficiently small so
that y + , y − ≥ 0. Then
Ay + = A(x + λw) = Ax + λAw = b + 0 = b,
Ay − = A(x − λw) = Ax − λAw = b − 0 = b,
so y + and y − are feasible, i.e., y + , y − ∈ K. But x = (y + + y − )/2 is a convex combination of
y + and y − , which contradicts the assumption that x is an extreme point of K. So Case 2 is
impossible.
(3 ) ⇒ (1 ). Suppose x is a BFS for the LP. We aim to show that x is a vertex of K,
that is, there exists c ∈ Rn such that cT x < cT y for all y ∈ P , y 6= x. Since x is a BFS, there
exists a set B ⊆ [n] with |B| = m such that the columns of AB are linearly independent,
AB xB = b, and xN = 0 (where N = [n] \ B). For j = 1, . . . , n, define
(
0, if j ∈ B;
cj =
1, if j ∈ N .
Note that cT x = cTB xB + cTN xN = 0T xB + cTN 0 = 0. For y ∈ K, we have y ≥ 0 (since y is
feasible), and clearly c ≥ 0, so cT y ≥ 0 = cT x. Furthermore, if cT y = 0, then yj must be 0
for all j ∈ N , so AB yB = b = AB xB . Multiplying on the left by A−1 −1
B gives yB = AB b = xB .
T
So x is the unique point in K for which c x = 0. Hence x is a vertex of K.
Definition 2.18. A polyhedron is pointed if it contains at least one vertex.
Note that a polyhedron contains a (bi-infinite) line if there exist vectors u, d ∈ Rn such
that u + λd ∈ K for all λ ∈ R.
Theorem 2.19. Let K ⊆ Rn be a polyhedron. Then K is pointed if and only if K does not
contain a (bi-infinite) line.
Note that the feasible region of an LP with nonnegativity constraints, such as an LP
in equational form, cannot contain a line. So this theorem shows (again) that every LP in
equational form that is feasible has a BFS (Fact 2.1).

2.4 Basic feasible solutions for general LPs

Note that we’ve defined basic feasible solutions for LPs in equational form, but not for
general LPs. Before we do that, let us make an observation about equational LPs, and the
number of tight constraints (i.e., those constraints that are satisifed at equality).
Consider an LP in equational form with n variables and m constraints, and let x be a
BFS. Then x satisfies all m equality constraints of the form ai x = bi . Since xN = 0, we see
that x additionally satisfies at least n−m nonnegativity constraints at equality. A constraint
is said to be tight if it is satisfied at equality, so we have the following fact.
LECTURE 2. GEOMETRY OF LPS 7

Fact 2.20. If x is a BFS of an LP in equational form with n variables, then x has at least
n tight constraints.

We can use this idea to extend the definition of BFS to LPs that are not in equational
form.

Definition 2.21. For a general LP with n variables, i.e., an LP of the form

min{ cT x | Ax ≥ b, x ∈ Rn },

a point x ∈ Rn is a basic feasible solution if x is feasible and there exist some n linearly
independent constraints that are tight (hold at equality) for x.

Proposition 2.22. For an LP in equational form, this definition of BFS and the previous
definition of BFS are equivalent.

(You may want to prove this for yourself.) Using this definition, one can now reprove
Theorem 2.17 for general LPs: i.e., show the equivalence between BFSs, vertices, and extreme
points holds not just for LPs in equational form, but for general LPs. We can use this fact
to find optimal solutions for LPs whose feasible regions are pointed polyhedra (and LPs in
equational form are one special case of this).