Elements of Convex Optimization Theory

Costis Skiadas

August 2015

This is a revised and extended version of Appendix A of Skiadas (2009), providing

a self-contained overview of elements of convex optimization theory. The presentation is
limited to the …nite-dimensional case, but from an advanced perspective that is amenable
to generalization. Some results, like the projection theorem and the Riesz representation
theorem for Hilbert spaces, are stated and proved in a way that also applies to in…nite-
dimensional spaces, since doing so comes at negligible additional cost. On the other hand,
instances where the …nite dimensionality assumption is essential are noted.

1 Vector Space
A (real) vector or linear space is a set X, whose elements are called vectors or points,
together with two operations, addition and multiplication by scalars, that satisfy the
following conditions, for all vectors x; y; z and real numbers ; :

1. To (x; y) addition assigns a unique element of X denoted by x + y; and to ( ; x)

multiplication by scalars assigns a unique element of X denoted by x:

2. x + y = y + x; x + (y + z) = (x + y) + z and ( x) = ( ) x:

3. (x + y) = x + y and ( + ) x = x + x:

4. 0 + x = x and 1x = x; where 0 is an element of X called the zero vector.

5. There exists a vector x such that x + ( x) = 0:

Kellogg School of Management, Department of Finance, Northwestern University, 2001 Sheridan Road,
Evanston, IL 60208.

1
It follows that the zero vector is unique, and for each x in X, the vector x is unique. We
write x y = x + ( y).
An underlying vector space X is taken as given throughout this appendix. Although the
term “vector space X”is common, it should be emphasized that a vector space speci…cation
includes not only a set of vectors but also the rules for adding vectors and multiplying by
scalars.

Example 1 X = Rd is the usual d-dimensional Euclidean space. Vector addition and

multiplication by scalars are de…ned coordinatewise: ( x + y)i = xi + yi for every scalar
; vectors x; y and coordinate i:

Example 2 Given an underlying probability space, the set of all random variables of …nite
variance is a vector space, with the usual rules for adding and scaling random variables
state by state.

A subset L of X is a linear (or vector) subspace if it is a vector space itself or, equiv-
alently, if for all x; y 2 L and 2 R; x + y and x are elements of L: The linear subspace
generated or spanned by a set of vectors S; denoted by span(S) ; is the intersection of
all linear subspaces that include S. Alternatively, the span of a set can be constructed
from within: For a …nite set of vectors S = fx1 ; : : : ; xn g ; the set span(S) ; also denoted
by span(x1 ; : : : ; xn ) ; consists of all linear combinations of the form 1 x1 + + n xn ; where
S
1 ; : : : ; n 2 R: For every set of vectors S; span(S) = fspan (F ) : F is a …nite subset of Sg :
A set of vectors S is linearly independent if every x 2 span(S) has a unique repre-
sentation of the form x = 1 x1 + + n xn ; for some 1 ; : : : ; n 2 R and x1 ; : : : ; xn 2 S:
It follows easily that the set of vectors S is linearly independent if and only if for all
x1 ; : : : ; xn 2 S and 1 ; : : : ; n 2 R;

1 x1 + + n xn = 0 implies 1 = = n = 0:

A basis of X is a linearly independent set of vectors S that generates X, that is, span(S) =
X.
A vector space is …nite dimensional if it has a …nite basis and in…nite dimensional
otherwise. Example 2 with an in…nite state space motivates our general interest in in…nite-
dimensional vector spaces. In this text, every state space is assumed to be …nite, and
therefore random variables can be encoded as …nite-dimensional Euclidean vectors. We
therefore focus on the …nite-dimensional case, but as noted in the introductory remarks,

2
from a perspective that allows direct extensions to contexts such as that of Example 2
with an in…nite state space. Every basis of a …nite-dimensional vector space has the same
number of elements, called the space’s dimension. The vector space f0g has, by de…nition,
dimension zero.
If X has …nite dimension d; we represent a basis fB1 ; : : : ; Bd g of X as a column matrix

B = (B1 ; : : : ; Bd )0 ;

and we write x to denote the row vector in Rd that represents the point x in X relative
to the given basis B, meaning that
d
X
x x
x= B= i Bi :
i=1

A functional is a function of the form f : X ! R. A functional f is linear if

f ( x + y) = f (x) + f (y) for all x; y 2 X and ; 2 R: Given a …nite basis B =
(B1 ; : : : ; Bd )0 and a linear functional f; we use the notation

f (B) = (f (B1 ) ; : : : ; f (Bd ))0

for the column matrix that lists the values that f assigns to the basis elements. The single
vector f (B) in Rd determines the entire function f; since x = x B implies f (x) = x f (B) :
A subset C of X is convex if for all x; y 2 C; 2 (0; 1) implies x + (1 ) y 2 C:
The function f : D ! R; where D X, is convex if D is a convex set and for all x; y 2 D;
2 (0; 1) implies f ( x + (1 ) y) f (x) + (1 ) f (y) : The function f is concave
if f is convex. A subset C of X is a cone if for all x 2 C; 2 R+ implies x 2 C: One
can easily check that a cone C is convex if and only if x; y 2 C implies x + y 2 C, and that
a convex cone C is a linear subspace if and only if x 2 C implies x 2 C:
An important type of convex set is a linear manifold. Given any subset S of X and
vector x; we write x + S = S + x = fx + s : s 2 Sg to denote the translation of S by x.
A subset M of X is a linear manifold if some translation of M is a linear subspace. The
dimension of a linear manifold M is that of the linear subspace to which M translates.
An exercise shows:

Proposition 3 The set M X is a linear manifold if and only if x; y 2 M and 2R

implies x + (1 ) y 2 M:

3
2 Inner Product
A widely applicable form of optimization can be thought of as a projection operation.
Informally, the projection of a point x onto a convex set S is the point of S that is closest
to x (assuming such a point exists). The simplest nontrivial instance of this problem is the
case in which S is a line. Suppose that S = span(y) and y has unit “length.” The scalar
s such that sy is the projection of x onto S is the inner product of x and y, denoted by
(y j x). Note that the functional (y j ) is linear. De…ning ( y j x) = (y j x) for all 2 R,
it is not hard to see that the function ( j ) is bilinear (linear in each argument), symmetric
((x j y) = (y j x)) and positive de…nite ((x j x) > 0 for x 6= 0). With this geometric
motivation, we take the notion of an inner product as a primitive object satisfying certain
axioms, and (in later sections) we use inner products to characterize projections on convex
sets more generally.

De…nition 4 A (real) inner product ( j ) on the vector space X is a mapping that

assigns to each (x; y) 2 X X a real number, denoted by (x j y), and satis…es, for all
x; y; z 2 X and 2 R;

1. (x + y j z) = (x j z) + (y j z) and ( x j y) = (x j y) :

2. (x j y) = (y j x) :

3. (x j x) 0; with equality holding if and only if x = 0.

An inner product space is a vector space together with an inner product on this space.

Example 5 Suppose X = Rd and Q is a positive de…nite symmetric matrix. Arranging

the elements of X as row vectors, the quadratic form (x j y) = xQy 0 de…nes an inner
product on X: If Q is the identity matrix, we obtain the Euclidean inner product:
P
(x j y) = x y = di=1 xi yi :

Example 6 Suppose X is any …nite-dimensional vector space and B is any basis of X:

Then an inner product is de…ned by letting (x j y) = x y ; where x = x B and y = y B:
We will later establish the interesting fact that given any inner product on X; we can select
the basis B so that this example’s representation is valid.

Example 7 Suppose X is the set of all random variables on some …nite probability space
in which every state is assigned a positive probability mass. Then (x j y) = E [xy] de…nes

4
an inner product. If Y is a linear subspace of X that does not contain the constant random
variables, then (x j y) = cov[x; y] = E [xy] ExEy de…nes an inner product on Y but not
on X (why?).

We henceforth take as given the inner product space (X; ( j )) ; which in the remainder of
this section is assumed to be …nite dimensional.
It will be convenient to extend the inner product notation to matrices of vectors, using
the usual matrix addition and multiplication rules. In particular, given a column matrix
of vectors B = (B1 ; : : : ; Bd )0 and any vector x; we write

x j B 0 = ((x j B1 ) ; : : : ; (x j Bd )) ;

and 0 1
(B1 j B1 ) (B1 j B2 ) (B1 j Bd )
B C
B (B2 j B1 ) (B2 j B2 ) (B2 j Bd ) C
BjB 0
=B
B .. .. .. .. C:
C
@ . . . . A
(Bd j B1 ) (Bd j B2 ) (Bd j Bd )
The matrix (B j B 0 ) is known as the Gram matrix of B and plays a crucial role in
the computation of projections. The following proposition shows that Gram matrices can
be used to convert abstract inner-product notation to concrete expressions involving only
matrices of scalars.

Proposition 8 Suppose B = (B1 ; : : : ; Bd )0 is a basis of X:

(a) For any x = x B and y = y B;
x
(x j y) = B j B0 y0
:

(b) The Gram matrix (B j B 0 ) is symmetric and positive de…nite.

(c) For any x 2 X; the representation x = x B can be computed as
x 1
= x j B0 B j B0 :

Proof. (a) Using the bilinearity of the inner product, we compute

x
(x j y) = B j B0 y0
= x
B j B0 y0
:

(b) The symmetry of the inner product implies that (B j B 0 ) is a symmetric matrix.
For any row vector 2 Rd , (B j B 0 ) 0 = ( B j B) : By the positive de…niteness of the

5
inner product it follows that (B j B 0 ) 0 0; with equality holding if and only if B = 0:
Since B is a basis, B = 0 if and only if = 0: This proves that (B j B 0 ) is a positive
de…nite matrix.
(c) By the linearity of the inner product, x = x B implies (x j B 0 ) = x (B j B 0 ) : By
part (b), (B j B 0 ) is invertible, and the claimed expression for x follows.

Two vectors x and y are orthogonal if (x j y) = 0. A set of vectors is orthogonal

if its elements are pairwise orthogonal. A basis B is orthonormal if it is orthogonal and
normalized so that (Bi j Bi ) = 1 for every i; or, equivalently, if the Gram matrix (B j B 0 )
is the identity matrix. Example 6 shows that every …nite basis is orthonormal under some
inner product. In Section 6 we will show that every …nite-dimensional inner-product space
has an orthonormal basis and can therefore be identi…ed with a Euclidean inner-product
space.
A Riesz representation of the linear functional f : X ! R is a vector x 2 X such
that f (y) = (x j y) for all y 2 X.

Proposition 9 Suppose B = (B1 ; : : : ; Bd )0 is a basis of X and f is a linear functional.

Then a Riesz representation of f exists, is unique and is given by
1
f (B)0 B j B 0 B:

Proof. For any x = xB and y = y B;

f (y) = f (B)0 y0
and (x j y) = x
B j B0 y0
:

The vector x is, therefore, a Riesz representation of f if and only if f (B)0 = x (B j B 0 ) :

Since (B j B 0 ) is positive de…nite, it is invertible, and therefore the last equality is equivalent
to x = f (B)0 (B j B 0 ) 1 :

3 Norm
A norm on the vector space X is a function of the form k k : X ! R that satis…es the
following properties for all x; y 2 X :

1. (triangle inequality) kx + yk kxk + kyk :

2. k xk = j j kxk for all 2 R.

6
 3. kxk 0; with equality holding if and only if x = 0.

Here kxk represents the value that the norm k k assigns to x, referred to simply as the
norm of x. It is easy to see that the triangle inequality holds for all x; y 2 X if and only
if
jkxk kykj kx yk for all x; y 2 X: (1)

In the current context, we assume that the norm k k is induced1 by the underlying
inner product ( j ) ; meaning that
p
kxk = (x j x); x 2 X: (2)

We will see shortly that k k so de…ned is indeed a norm. We think of kxk as the length of
x, in the sense used in our earlier informal motivation of inner products. Orthogonality of
two vectors x and y, de…ned by the condition (x j y) = 0; can be characterized entirely in
terms of the induced norm.

Proposition 10 The vectors x and y are orthogonal if and only if they satisfy the Pythagorean
identity
kx + yk2 = kxk2 + kyk2 :

Proof. The claim follows from the computation

kx + yk2 = (x + y j x + y) = (x j x) + (y j y) + 2 (x j y) = kxk2 + kyk2 + 2 (x j y) :

Two vectors x and y are said to be colinear if either x = y or y = x is true for some
2 R:

Proposition 11 Equation (2) de…nes a norm k k that satis…es the Cauchy-Schwarz

inequality
j(x j y)j kxk kyk ; x; y 2 X;

with equality holding if and only if x and y are colinear.

1
Every norm we use in this text is induced by some inner product. Not every norm can be induced by
an inner product, however; a norm that is not translation invariant cannot be induced by an inner product.

7
 Proof. The Cauchy-Schwarz inequality holds trivially as an equality if either x or y is
zero. Suppose x and y are nonzero, and let x ^ = x= kxk and y^ = y= kyk. Visualizing the
vector (^
x j y^) y^ as the projection of x
^ on the line spanned by y^, we note that x
^ (^
x j y^) y^
is orthogonal to y^: Indeed,

(^
x (^
x j y^) y^ j y^) = (^
x j y^) (^
x j y^) (^
y j y^) = 0:

The Pythagorean identity then implies that

0 k^
x x j y^) y^k2 = 1
(^ x j y^)2 ;
(^

which implies the Cauchy-Schwarz inequality. Equality holds if and only if x

^ = (^
x j y^) y^, a
condition that is equivalent to x = y for some scalar . We still must verify that k k is a
norm. We use the Cauchy-Schwarz inequality to show the triangle inequality:

kx + yk2 = kxk2 + 2 (x j y) + kyk2

kxk2 + 2 kxk kyk + kyk2 = (kxk + kyk)2 :

The remaining norm properties are immediate from the de…nitions.

4 Convergence
A sequence fxn g = (x1 ; x2 ; : : : ) of points in X converges to the limit x 2 X if for every
" > 0; there exists an integer N" such that n > N" implies kxn xk < ". In this case, the
sequence fxn g is said to be convergent. A subset S of X is closed if every convergent
sequence in S converges to a point in S: A subset S of X is open if its complement X n S
is closed. The set of all open subsets of X is known as the topology of X:
The following properties of closed and open sets can be easily con…rmed.2 The empty
set and X are both open and closed. The union of …nitely many closed sets is closed, and
the intersection of …nitely many open sets is open. Arbitrary intersections of closed sets
are closed, and arbitrary unions of open sets are open.
A sequence fxn g of points in X is Cauchy if for every " > 0; there exists an integer N"
such that m; n > N" implies kxm xn k < ": A subset S of X is complete if every Cauchy
2
Just as a norm need not be induced by an inner product, a topology, de…ned as a set of subsets of X
with the stated properties, need not be de…ned by a norm in a more general theory of convergence, which
is not needed in this text.

8
sequence in S converges to a limit in S. It is immediate from the de…nitions that a subset
of a complete set is complete if and only if it is closed. The triangle inequality implies
that every convergent sequence is Cauchy. Exercise 4 shows that the converse need not be
true for an arbitrary inner product space. Intuitively, a Cauchy sequence should converge
to something, but if that something is not within the space X; then the sequence is not
convergent. As we will see shortly, di¢ culties of the sort do not arise in …nite-dimensional
spaces.
The following proposition shows that if X is …nite dimensional, the convergence or
Cauchy property of a sequence is equivalent to the respective property of the sequence’s
coordinates relative to any given basis.

Proposition 12 Suppose B = (B1 ; : : : ; Bd )0 is a basis of X; and n = 1n ; : : : ; dn 2 Rd

for n = 1; 2; : : : The sequence f n Bg is Cauchy (resp. converges to B) if and only if the
scalar sequence i
n is Cauchy (resp. converges to i ) for every coordinate i 2 f1; : : : ; dg :

Proof. Suppose in is Cauchy for each i: The triangle inequality implies k n B m Bk

P i i kB k ; from which it follows that f
i n m i n Bg is Cauchy. Conversely, suppose
f n Bg is Cauchy. Noting the identity n = ( n B j y) ; where y = B 0 (B j B 0 ) 1 , we apply
the Cauchy-Schwarz inequality, to obtain in i = j(
m nB m B j yi )j k nB m Bk kyi k :
Therefore i
n is Cauchy. The claims in parentheses follow by the same argument, with
in place of m :

A Hilbert space is any inner product space that is complete (relative to the norm
induced by the inner product). One of the fundamental properties of the real line is that
it is complete. Given this fact, the last proposition implies:

Proposition 13 Every …nite-dimensional inner product space is a Hilbert space.

Another consequence of Proposition 12 is that for a …nite-dimensional vector space,

the convergence or Cauchy property of a sequence does not depend on the choice of an
underlying inner product, and therefore the topology on a …nite-dimensional vector space is
invariant to the inner product used to de…ne it (or the norm used to de…ne it, as explained
in Section 8). The situation is radically di¤erent for in…nite-dimensional spaces, where
di¤erent norms can de…ne di¤erent topologies.
We close this section with some topological terminology and notation that is of frequent
use. Given any r 2 R+ and x 2 X; the open ball with center x and radius r is the set
B (x; r) = fy 2 X : ky xk < rg :

9
Let S be any subset of X: The closure of S is the set

S = fx : B (x; ") \ S 6= ; for all " > 0g ;

the interior of S is the set

S 0 = fx : B (x; ") S for some " > 0g ;

and the boundary of S is the set S n S 0 : A vector x is a closure point of the set S if and
only if there exists a sequence in S that converges to x: Therefore, the set S is closed if and
only if S = S; and it is open if and only if S = S 0 : Finally, note that closures and interiors
can be described in purely topological terms: The closure of a set is the intersection of all
its closed supersets, and the interior of a set is the union of all its open subsets.

5 Continuity
A function f : D ! R; where D X; is continuous at x 2 D if for any sequence
fxn g in D converging to x, the sequence ff (xn )g converges to f (x) : The function f is
continuous if it is continuous at every point of its domain D: It is straightforward to check
that f : D ! R is continuous at x 2 D if and only if given any " > 0; there exists some
> 0 (depending on ") such that y 2 D \ B (x; ) implies jf (y) f (x)j < ": Based on last
section’s discussion, we note that if X is …nite dimensional, the continuity of f at a point
is true or not independently of the inner product (or norm) used to de…ne the topology.
Inequality (1) shows that the underlying norm is a continuous function. The inner
product ( j ) is also continuous, in the following sense.

Proposition 14 Suppose fxn g and fyn g are sequences in X converging to x and y, re-
spectively. Then f(xn j yn )g converges to (x j y) :

Proof. By the triangle and Cauchy-Schwarz inequalities, we have

j(xn j yn ) (x j y)j = j(x xn j y yn ) + (xn x j y) + (x j yn y)j

kx xn k ky yn k + kxn xk kyk + kxk kyn yk :

Letting n go to in…nity completes the proof.

Exercise 5 gives an example of a linear functional on an in…nite-dimensional space that

is not continuous. In a …nite-dimensional space, however, if f is a linear functional on X;

10
then it has a Riesz representation z (by Proposition 9). An application of the Cauchy-
Schwarz inequality shows that jf (x) f (y)j kzk kx yk for all x; y 2 X and hence the
continuity of f: The remainder of this section extends this argument to concave functions,
showing in particular that in a …nite-dimensional space, a concave function over an open
domain is continuous.
Consider any concave function f : D ! R; where D is an open subset of X: We
introduce two properties of f; which turn out to be equivalent to the continuity of f: We
say that f is locally bounded below if given any x 2 D; there exists a small enough
radius r > 0 such that the in…mum of f over the ball B (x; r) is …nite. We say that f is
locally Lipschitz continuous if given any x 2 D; there exist a constant K and a ball
B (x; r) D such that

jf (y) f (x)j K ky xk for all y 2 B (x; r) : (3)

Clearly, if f is locally Lipschitz continuous, it is continuous and therefore locally bounded

below. The following lemma shows that the converse implications are also true.

Lemma 15 Suppose that D X is open and convex, and the function f : D ! R is

concave. If f is locally bounded below, then f is locally Lipschitz continuous.

Proof. Given any x 2 D; let b denote a lower bound of f on B (x; r) D: Fixing any
y 2 B (x; r) ; let u = (y x) = ky xk and ( ) = f (x + u) for 2 [ r; r] : Consider the
following claim, where the …rst equality de…nes K :

(0) b (0) ( r) (ky xk) (0)

K=
r r ky xk
(r) (0) b (0)
= K:
r r
The function : [ r; r] ! R is concave and bounded below by b; which justi…es the …rst
and last inequalities. The three middle expressions represent slopes that decrease from left
to right since is concave. This proves (3) :

Remark 16 As in the lemma, assume that D is convex and open and f : D ! R is

concave. Exercise 3 shows that f is locally bounded below if it is bounded below over some
open ball within its domain. A corollary of this is that f is locally Lipschitz continuous if
it is continuous at a single point of its domain.

11
 The preceding lemma (and remark) applies to an arbitrary inner product space X:
Further restricting X to be …nite-dimensional, we can verify that a concave function is
locally bounded below, thus proving its continuity.

Theorem 17 Suppose that X is …nite dimensional, D is an open convex subset of X and

f : D ! R is concave. Then f is locally Lipschitz continuous, and therefore continuous.

Proof. Suppose …rst that X = Rd with the standard Euclidean inner product. We …x
any x 2 D and show that f is bounded below near x: An exercise shows that x is contained
in the interior of a set of the form

[ ; ] = fx 2 X : xi 2 [ i ; i] ; i = 1; : : : ; dg D;

where ; 2 Rd and i < i for all i: Since f is concave, it is minimized over [ ; ] at some
extreme point of [ ; ] ; that is, a point x such that xi 2 f i ; i g for all i: To see why, take
any x 2 [ ; ] and any k 2 f1; : : : ; dg ; and de…ne the points x and x by xi = xi = xi
for i 6= k, xk = k and xk = k : Then for some 2 [0; 1] ; x = x + (1 ) x : Concavity
implies min f (x ) ; f x f (x) : We can therefore replace x by one of x or x without
increasing its value under f: Repeating this process for all coordinates shows that for every
x 2 [ ; ] ; there exists some extreme point x of [ ; ] such that f (x) f (x) : Since [ ; ]
has only …nitely many extreme points, f is bounded below on [ ; ] : Continuity of f at x
follows from Lemma 15.
Finally, suppose X is an arbitrary …nite-dimensional space and let B be any basis. We
have seen in Proposition 12 that whether f is continuous or not does not depend on what
inner product we endow X with. We can therefore choose the inner product of Example 6,
which makes B an orthonormal basis. Identifying each element x = x B 2 X with its
basis representation x ; the above argument applies.

6 Projections on Convex Sets

Projections were introduced informally in motivating inner products. This section de-
…nes projections formally, shows the existence of projections on complete convex sets, and
establishes a dual characterization of projections on convex sets in terms of inner products.

De…nition 18 Let S be any subset of X. The vector xS is a projection of x on the set

S if xS 2 S and kx sk kx xS k for all s 2 S: The vector z 2 X supports the set S
at s 2 S if (z j s s) 0 for all s 2 S.

12
 The inequality (z j s s) 0 can be visualized as the requirement that the vectors z
and s s form an acute angle. Theorem 19(a) below characterizes projections on a convex
set S based on the intuitive geometric idea that a point xS in S is the projection on S of
a point x outside S if and only if for any s 2 S the vectors xS x and s xS form an
acute angle. The theorem also addresses the uniqueness and existence of a projection on
a convex set. Clearly, if S is not closed then a projection of a point on S may not exist.
For example, on the real line, the projection of zero on the open interval (1; 2) does not
exist. Exercise 4 further shows that in an in…nite-dimensional space, S can be closed and
convex, and yet the projection of zero on S may still not exist. The key issue in this case
is the absence of completeness of the ambient space X:
The central result on projections on convex sets follows. The theorem and its proof
apply to any inner product space (not necessarily a …nite-dimensional one).

Theorem 19 (Projection Theorem) Suppose that S is a convex subset of X and x; y 2

X:
(a) The vector xS is a projection of x on S if and only if xS 2 S and xS x supports
S at xS .
(b) Suppose xS is a projection of x on S and yS is a projection of y on S: Then
kxS yS k kx yk :
(c) There exists at most one projection of x on S:
(d) If S is complete, then the projection of x on S exists.

Proof. (a) Suppose xS x supports S at xS 2 S. Then for any s 2 S,

kx sk2 = kx xS k2 + ks xS k2 + 2 (xS xjs xS ) kx xS k2 ;

proving that xS is a projection of x on S. Conversely, suppose xS is a projection of x on

S. Given any nonzero s 2 S and 2 (0; 1), de…ne x = xS + (s xS ) 2 S. Then
1
2 (xS xjs xS ) = ks xS k2 + kx x k2 kx xS k2

ks xS k2 :

Letting approach zero, we conclude that (xS x j s xS ) 0.

(b) Let = y x and S = yS xS . By part (a), (xS x j S ) 0 and (y yS j S)
0. Adding the two inequalities, we obtain ( S j S) 0: Finally,

k k2 = k Sk
2
+k Sk
2
+ 2( S j S) k 2
Sk :

13
 (c) Apply part (b) with x = y.
(d) Let = inf fkx sk : s 2 Sg, and choose a sequence fsn g in S such that fkx sn kg
converges to : Direct computation (see Exercise 2) shows
2
sm + sn
ksm sn k2 = 2 kx sm k2 + 2 kx s n k2 4 x
2
2 kx sm k2 + 2 kx s n k2 4 2;

implying ksm sn k2 ! 0 as m; n ! 1: Therefore, the sequence fsm g, being Cauchy,

converges to some vector xS in S, since S is assumed complete. By continuity of the norm
operator it follows that = kx xS k and therefore xS is a projection of x on S.

Corollary 20 If X is a Hilbert space and S is a closed convex subset of X; the projection

of any x 2 X on S exists.

7 Orthogonal Projections
This section elaborates on the important special case of projections on linear manifolds.
A vector x is orthogonal to the linear manifold M if x is orthogonal to m1 m2 for all
m1 ; m2 2 M: The orthogonal to M subspace, denoted by M ? ; is the linear subspace
of all vectors that are orthogonal to M . Note that M ? = (x + M )? for every x 2 X; and
a vector supports M at some point if and only if it is orthogonal to M . Theorem 19(a)
applied to linear manifolds reduces to the following result.

Corollary 21 (orthogonal projections) Suppose M is a linear manifold in X and x 2

X: A point xM is the projection of x on M if and only if xM 2 M and x xM 2 M ? :

Projections on …nite-dimensional subspaces can be expressed by simple formulas in

terms of a given basis. Note that the following proposition, which generalizes Proposi-
tion 8(c), does not assume that the ambient space X is …nite dimensional.

Proposition 22 Suppose L is a linear subspace of X; and B = (B1 ; : : : ; Bd )0 is a basis of

L: Given any x 2 X; the vector
1
xL = x j B 0 B j B0 B (4)

is the projection of x on L; and x xL is the projection of x on

L? = fy 2 X : (B j y) = 0g : (5)

14
Finally, L?? = L:

1
Proof. Let xL = xL B; where xL = (x j B 0 ) (B j B 0 ) . Clearly, xL 2 L: Moreover,
x xL 2 L? , since, for any y = y B 2 L;

xL 1
(xL j y) = B j B0 y0
= x j B0 B j B0 B j B0 y0
= (x j y) :

By Corollary 21, xL is the projection of x on L: It is immediate that L L?? . Therefore,

x (x xL ) = xL 2 L?? and x xL 2 L? . Again by Corollary 21, it follows that x xL
is the projection of x on L? : Finally, we show that L?? L: Consider any x 2 L?? and
let xL be its projection on L; given by (4). We can then write x = xL + n; where n 2 L? :
Since (x j n) = (xL j n) + (n j n) and (x j n) = (xL j n) = 0; it follows that (n j n) = 0 and
therefore x = xL 2 L:

Corollary 23 Suppose that B = (B1 ; : : : ; Bd )0 is a column matrix of linearly independent

vectors in X; and M = fy 2 X : (B j y) = bg for some column vector b 2 Rd : Then M ? =
span(B1 ; : : : ; Bd ) ; and the projection of x on M exists and is given by
1
xM = x x j B0 b0 B j B0 B:

Proof. Fixing any m 2 M; note that M m is the linear subspace L? of equation (5) ;
and therefore M ? = L?? = L = span(B) : The point xM is the projection of x on M if and
only if xM m is the projection of x m on L? : We can therefore apply Proposition (22)
to conclude that the projection xM of x on M exists and
1
xM m = (x m) x m j B0 B j B0 B:

Since (B j m) = b; the claimed expression for xM follows.

Corollary 24 Suppose M is a …nite-dimensional linear manifold in X. Then every vector

x 2 X has a unique decomposition of the form x = m + n; where m 2 M and n 2 M ? :

The preceding corollary is not valid if the assumption that M is …nite-dimensional is

relaxed, in which case M need not be closed. The result extends easily to any Hilbert
space, however, still as a corollary of the Projection Theorem, provided the vector m is
only required to lie in the closure of M (rather than M ).

15
 If the linear subspace L has a …nite orthogonal basis B; then (B j B 0 ) is diagonal and
formula (4) for the projection of x on L reduces to
k
X (x j Bi )
xL = Bi : (6)
(Bi j Bi )
i=1

Assuming X is …nite-dimensional, this equation can be used to recursively construct an

orthogonal basis of X; a process known as Gram-Schmidt orthogonalization. We start
with any nonzero vector B1 in X: Assuming we have constructed k orthogonal vectors
B1 ; : : : ; Bk such that L = span(B1 ; : : : ; Bk ) 6= X; we select any x 2 X n L and de…ne
Bk+1 = x xL ; where xL is the projection of x on L, given by (6). If X is d-dimensional,
the recursive construction terminates for k = d: Normalizing the resulting orthogonal basis
proves

Proposition 25 Every …nite-dimensional inner product space has an orthonormal basis.

Formula (4) is the same as the Riesz representation expression of Proposition 9 with
f (y) = (x j y) : The reason should be clear given the following general relationship be-
tween orthogonal projections and Riesz representations (whose proof is immediate from
the de…nitions).

Proposition 26 Suppose that f (y) = (x j y) for all y 2 X and fL is the restriction of f

on the linear subspace L. The vector xL is the Riesz representation of fL in L if and only
if it is the projection of x on L:

We use orthogonal projections to show that in any Hilbert space the Riesz represen-
tation of a continuous linear functional exists. The argument is redundant in the …nite-
dimensional case, since the claim was established in Proposition 9, but still worth reviewing.

Theorem 27 In a Hilbert space, a linear functional is continuous if and only if it has a

(necessarily unique) Riesz representation.

Proof. Suppose f is a continuous linear functional. The null subspace N = fx : f (x) = 0g

is closed. If X = N , the result follows trivially. Otherwise there exists z 2 X such that
f (z) 6= 0. Let zN be the projection of z on N , and de…ne y = f (z zN ) 1 (z zN ) : Then
y is orthogonal to N and satis…es f (y) = 1. For any x 2 X; the fact that x f (x) y 2 N
implies that (x f (x) y j y) = 0; and therefore f (x) = (x j y) = (y j y) : The normalized
vector (y j y) 1 y is the Riesz representation of f:

16
8 Compactness
In this section we show that a continuous function achieves a maximum and a minimum
over any compact set. While the result applies to any normed space, it is mostly useful in
the …nite-dimensional case, in which a set is compact if and only if it is closed and bounded.
A subset S of X is (sequentially) compact if every sequence in S has a subsequence
that converges to a vector in S: This de…nition immediately implies that a compact set
is complete and therefore closed. A compact set S is also bounded, meaning that
sup fksk : s 2 Sg < 1: (If S were unbounded, there would exist a sequence fsn g in S
such that ksn k > n for all n; a condition that precludes the existence of a convergent
subsequence.)

Proposition 28 In a …nite-dimensional space, a set is compact if and only if it is closed

and bounded.

Proof. As noted above, it is generally true that a compact set is closed and bounded.
We prove the converse, relying on the assumption that X is …nite dimensional. By
Proposition 12 and Example 6, it su¢ ces to show the result for X = Rd (why?). Let
fsn g = s1n ; : : : ; sdn be a sequence in a closed bounded subset S of Rd . Then the …rst
coordinate sequence s1n lies in some bounded interval I. Select a half-interval I1 of I
that contains in…nitely many points of s1n and let s1n1 be one of these points. Then select
a half-interval I 2 of I 1 that contains in…nitely many points of s1n : n > n1 and let s1n2 be
one of these points. Continuing in this manner, we obtain a nested sequence of intervals
fIk g whose length shrinks to zero and a corresponding subsequence s1nk with s1nk 2 Ik
for all k. Clearly, the subsequence s1nk is Cauchy and therefore convergent. Repeating
the argument we can extract a further subsequence for the second coordinate, then the
third, and so on. This process generates a convergent subsequence of fsn g whose limit
point must be in S, since S is assumed closed.

The preceding proposition is not valid in in…nite-dimensional spaces. It is not hard to

prove that a set S in any normed space is compact if and only if it is complete and for
every " > 0 there exists a …nite collection of balls of radius " whose union contains S: This
generalization will not be useful in our applications, however— a more fruitful direction is
brie‡y discussed at the end of this section.

17
Proposition 29 Suppose that S is a compact subset of X and the function f : S ! R is
continuous. Then there exist s ; s 2 S such that

f (s ) f (s) f (s ) for all s 2 S:

Proof. Let fsn g be a sequence such that limn f (sn ) = sup f: By the compactness of
S, there exists a subsequence of fsn g converging to some s 2 S. Since f is continuous,
f (s ) = sup f and therefore f (s ) f (s) for all s 2 S. The same argument applied to
f completes the proof.

Proposition 30 Suppose X is …nite-dimensional, A and B are closed subsets of X and at

least one of them is bounded. Then there exists a pair (a; b) 2 A B such that ka bk
kx yk for all (x; y) 2 A B:

Proof. Suppose A is bounded and therefore compact. Select r > 0 large enough so
that the set C = B \ fy : kx yk r for some x 2 Ag is nonempty. Note that C is also
compact (why?). Let f(an ; bn )g be any sequence in A C such that kan bn k converges
to = inf fkx yk : (x; y) 2 A Cg : We can then extract a subsequence f(ank ; bnk )g that
converges to some (a; b) 2 A C: By the triangle inequality,

ka bk ka ank k + kank bnk k + kbnk bk :

The right-hand side converges to ; and therefore ka bk = : For any (x; y) 2 A B;

either y 2 C, in which case kx yk by the de…nition of ; or y 2 = C; in which case
kx yk > r by the de…nition of C: Since r ; the proof is complete.

Proposition 29 implies the equivalence of all norms on a …nite-dimensional space, in

the following sense. Suppose X is …nite dimensional and consider any other norm k k on
X (not necessarily induced by any inner product). Since k k is convex, it is continuous
on the open ball B (0; 2) and therefore achieves a minimum and a maximum over the
closure of B (0; 1) : By the homogeneity of k k , there exist constants k and K such that
k kxk kxk K kxk for all x 2 X: Therefore, for a …nite-dimensional vector space, all
norms de…ne the same topology and the same compact sets.
In in…nite-dimensional spaces, closed and bounded sets are not compact and di¤erent
norms can de…ne dramatically di¤erent topologies. This severely limits the usefulness of
compactness, in this section’s sense, for in…nite-dimensional optimization theory. A partial
remedy is provided by the notion of weak compactness, which is particularly useful in

18
conjunction with convexity properties. For example, a weak compactness argument can
be used to prove that in any Hilbert space, if the nonempty set S is convex, closed and
bounded, and the function f : X ! R is concave over S and continuous at every point of
S; then f achieves a maximum in S: Suitable references are provided in the endnotes.

9 Supporting Hyperplanes
A hyperplane H is a linear manifold whose orthogonal subspace is of dimension one. If
H ? is spanned by the vector y and = (y j x) ; where x is any point in H; then

H = fx : (y j x) = g : (7)

This expression characterizes all hyperplanes as y ranges over the set of all nonzero vectors
and ranges over the set of all scalars.
According to De…nition 18, the vector y supports the set S X at s 2 S if and only if
(y j s) (y j s) for all s 2 S. Letting = (y j s) in (7) ; this condition can be visualized
as S being included in the half-space fx : (y j x) g on the one side of the hyperplane
H; while touching H at s: The following is an extension of De…nition 18, which can be
visualized the same way, but does not require s to be an element of S:

De…nition 31 The vector y 2 X supports the set S at s 2 X if

(y j s) = inf f(y j s) : s 2 Sg : (8)

It is intuitively compelling that one should be able to support a convex set at any
point of their boundary. The supporting hyperplane theorem, proved below, shows that
this is indeed true for a …nite-dimensional space. In another example in which …nite-
dimensional intuition need not apply in in…nite-dimensional spaces, Exercise 16 shows that
in an in…nite-dimensional space it may not be possible to support a convex set at a point
of its boundary.3

Theorem 32 (Supporting Hyperplane Theorem) Suppose S is a convex subset of

the …nite-dimensional inner-product space X: Then, given any vector x that is not in the
interior of S; there exists a nonzero vector y such that (y j x) (y j s) for all s 2 S: If x
is on the boundary of S; then y supports S at x:
3
The supporting hyperplane theorem does stay true if the assumption of a …nite-dimensional X is
replaced by the assumption that S has an interior point. But, as in the example of Exercise 16, this is
typically too restrictive an assumption for applications.

19
 Proof. Since x is not interior, we can construct a sequence fxn g of vectors such that
xn 2= S for all n and xn ! x as n ! 1: For each n, let sn be the projection of xn on
S and de…ne yn = (sn xn ) = ksn xn k. The dual characterization of projections gives
(yn j sn ) (yn j s) for all s 2 S. By Theorem 19(b), the sequence fsn g converges to the
projection s of x on S. The sequence fyn g lies in the closure of the unit ball, which is
compact, and therefore we can extract a subsequence fynk g that converges to some vector
y of unit norm. By the continuity of inner products, we conclude that (y j s) (y j s) for
all s 2 S. If x is on the boundary of S, then x = s is the limit of some sequence fsn g
in S. Therefore, f(y j sn )g converges to (y j s), implying (8) : If x is not on the boundary
of S; then fynk g converges to y = (s x) = ks xk. Since (y j s x) > 0, it follows that
(y j x) < (y j s) (y j s) for all s 2 S:

Corollary 33 (Separating Hyperplane Theorem) Suppose A and B are convex sub-

sets of the …nite-dimensional inner-product space X: If A \ B = ;; then there exists a
nonzero y 2 X such that
inf (y j a) sup (y j b) : (9)
a2A b2B

Proof. Let S = A B = fa b : a 2 A; b 2 Bg : Since A \ B = ;; zero is not in S.

If X is in…nite-dimensional, the fact that at least one of A and B has a nonempty interior
implies that S has a nonempty interior. By the supporting hyperplane theorem, there
exists a nonzero y 2 X such that (y j s) 0 for all s 2 S and therefore (y j a) (y j b) for
every a 2 A and b 2 B:

10 Superdi¤erential and Gradient

Consider any function f : D ! R; where D X: The vector y 2 X is a supergradient
of f at x 2 D if it satis…es the gradient inequality:

f (x + h) f (x) + (y j h) for all h such that x + h 2 D:

The superdi¤erential of f at x; denoted by @f (x) ; is the set of all supergradients of f

at x:
The supergradient property can be visualized as a support condition in the space X R
with the inner product

((x1 ; 1) j (x2 ; 2 )) = (x1 j x2 ) + 1 2; xi 2 X; i 2 R:

20
The subgraph of f is the set

sub (f ) = f(x; ) 2 D R: f (x)g : (10)

The preceding de…nitions imply that

y 2 @f (x) () (y; 1) supports sub (f ) at (x; f (x)) : (11)

Theorem 34 In a …nite-dimensional inner product space,4 the superdi¤ erential of a con-

cave function at an interior point of its domain is nonempty, convex and compact.

Proof. Suppose f : D ! R is concave and x 2 D0 : By the supporting hyperplane

theorem (Theorem 32), sub(f ) is supported by some nonzero (y; ) 2 X R at (x; f (x)) :

(y j x) f (x) = min f(y j z) : f (z) ; z 2 D; 2 Rg :

Since the left-hand side is …nite, it follows that 0: If = 0; then y supports D at x,

which contradicts the assumption that x is an interior point of D: Therefore, > 0 and
y= 2 @f (x) : This proves that @f (x) is nonempty. It follows easily from the de…nitions
that @f (x) is also convex and closed. Finally, we show that @f (x) is bounded, utilizing
the …nite dimensionality assumption. We assume x = 0; which entails no loss of generality
(why?). Theorem 17 implies that f is bounded below over some ball centered at zero. Let
" > 0 and K 2 R be such that for all z 2 X; kzk = " implies z 2 D0 and f (z) > K: It
follows that for every y 2 @f (0) ;

" "
K<f y f (0) + y j y = f (0) " kyk ;
kyk kyk

which results in the bound kyk < (f (0) K) =": This proves that @f (x) bounded. Since
it is also closed, it is compact.

Assuming it exists, the directional derivative of f at x in the direction h is

denoted and de…ned by
f (x + h) f (x)
f 0 (x; h) = lim :
#0

The gradient of f at x is said to exist if f 0 (x; h) exists for every h 2 X and the functional
f 0 (x; ) is linear. In this case, the Riesz representation of the linear functional f 0 (x; ) is
4
The result extends to general Hilbert spaces, provided the compactness conclusion is replaced by weak
compactness.

21
the gradient of f at x; denoted by rf (x) : Therefore, when it exists, the gradient rf (x)
is characterized by the restriction

f 0 (x; h) = (rf (x) j h) ; for all h 2 X:

For concave functions, directional derivatives and gradients are related to superdi¤er-
entials as follows.

Theorem 35 Suppose D is a convex subset of the …nite-dimensional inner product space

X; the function f : D ! R is concave, and x is any interior point of D: Then the directional
derivative f 0 (x; h) exists for all h 2 X; and the following three claims are true:
(a) @f (x) = fy 2 X j (y j h) f 0 (x; h) for all h 2 Xg :
(b) f 0 (x; h) = min f(y j h) : y 2 @f (x)g for every h 2 X:
(c) The gradient rf (x) exists if and only if @f (x) is a singleton, in which case @f (x) =
frf (x)g :

Proof. Fixing any h 2 X; let A = 2 R : x + h 2 D0 (an open interval) and

de…ne the concave function : A ! R by ( ) = f (x + h) f (x) : Note that f 0 (x; h) =
0
+ (0) (the right derivative at zero). For each 2 A; let ( ) = ( ) = ; which is
the slope of the line segment on the plane connecting the origin to the point ( ; ( )) :
Consider any decreasing sequence f n g of positive scalars in A: Concavity of implies
that the corresponding sequence of slopes f ( n )g is increasing. Moreover, …xing any
" > 0 such that x "h 2 D0 ; we have ( n ) ( ") for all n: Being increasing
and bounded, the sequence f ( n )g converges. This proves that the (monotone) limit
f 0 (x; h) = lim #0 ( ) exists and is …nite.
(a) If y 2 @f (x) ; the gradient inequality implies that (y j h) ( ) for all positive
0
2 A: Letting # 0; it follows that (y j h) f (x; h) : Conversely, suppose that y 2 = @f (x) ;
and therefore (1) = f (x + h) f (x) > (y j h) for some h: We saw earlier that ( )
increases to f 0 (x; h) as # 0; and therefore f 0 (x; h) (1) > (y j h) for some h:
(b) In light of part (a), the claim of part (b) follows if given any h 2 X; we can produce
0
some y 2 @f (x) such that (y j h) = f 0 (x; h) : The gradient inequality ( ) + (0) for
0
all 2 A (equivalently, f (x + h) f (x) + f (x; h) for all 2 A), can be restated as
the condition that the line segment

L= x + h; f (x) + f 0 (x; h) : 2A X R

22
does not intersect the interior of the subgraph of f; as de…ned in (10) : By the separating
hyperplane theorem (Corollary 33), there exists nonzero (p; ) 2 X R that separates the
sets L and sub(f ) :

inf (p j x + h) + f (x) + f 0 (x; h) sup (p j x

~) + :
2A x; )2sub(f )
(~

Since the right-hand side must be …nite, > 0: It follows that the right-hand side is at
least as large as (p j x) + f (x) ; which is also obtained as the expression on the left-hand
side with = 0: Since 0 2 A0 ; the coe¢ cient of on the left-hand side must vanish:
(p j h) + f 0 (x; h) = 0: Therefore, f 0 (x; h) = (y j h) ; where y = p= ; and the separation
condition reduces to the gradient inequality de…ning the condition y 2 @f (x) :
(c) If @f (x) = f g ; then part (b) implies that f 0 (x; h) = ( j h) for all h 2 X;
and therefore = rf (x) : Conversely, if the gradient exists, then part (a) implies that
(rf (x) y j h) 0 for all h 2 X and y 2 @f (x) : Letting h = rf (x) y; it follows that
y = rf (x) if y 2 @f (x) :

11 Global Optimality Conditions

We take as given the set C X; and the functions U : C ! R and G : C ! Rn ; for some
positive integer n: For each 2 Rn ; we consider the constrained optimization problem

J ( ) = sup fU (x) : G (x) ; x 2 Cg : (P )

With the convention inf ; = 1; this de…nes a function of the form J : Rn ! [ 1; +1] ;
which is clearly monotone: 1 2 =) J ( 1 ) J ( 2 ) : We will characterize the solution
to the problem (P0 ) : Since problem (P ) is the same as problem (P0 ) with G in place
of G; this covers the general case. The key to understanding (P0 ), however, is to consider
the entire function J:
Associated with problem (P0 ) is the Lagrangian

L (x; ) = U (x) G (x) ; x 2 C; 2 Rn ;

where the dot denotes the Euclidean inner product in Rn : The parameter will be referred
to as a Lagrange multiplier. Assuming J (0) is …nite, we extend our earlier de…nition
(for …nite-valued functions) by de…ning the superdi¤erential of J at zero to be the set

@J (0) = f 2 Rn : J ( ) J (0) for all 2 Rn g :

23
The monotonicity of J implies that @J (0) Rn+ : Indeed, if 2 @J (0) ; then 0 J ( )
J (0) for all 0 and therefore 0: The following relationship between the
Lagrangian and the superdi¤erential of J at zero is key.

Lemma 36 Suppose that J (0) is …nite. Then for every 2 Rn+ ;

2 @J (0) () J (0) = sup fL (x; ) : x 2 Cg :

Proof. In the space Rn R with the Euclidean inner product, 2 @J (0) if and only
if ( ; 1) supports at (0; J (0)) (in the sense of De…nition 31) the set

S1 = f( ; ) 2 Rn R: < J ( )g :

Similarly, J (0) = sup fL (x; ) : x 2 Cg if and only if ( ; 1) supports at (0; J (0)) the set

S2 = f( ; ) 2 Rn R : G (x) and < U (x) for some x 2 Cg :

The proof is concluded by noting that S1 = S2 :

Lemma 36 does not assume that an optimum is achieved. If a maximum does exist,
the argument can be extended to obtain the following global optimality conditions for
problem (P0 ) :

Theorem 37 Suppose that c 2 C; G (c) 0 and 2 Rn : Then the following two conditions
are equivalent:

1. U (c) = J (0) and 2 @J (0) :

2. L (c; ) = maxx2C L (x; ) ; G (c) = 0; 0:

Proof. (1 =) 2) Suppose condition 1 holds. We noted that 2 @J (0) implies

0: By the last lemma, U (c) = sup fL (x; ) : x 2 Cg : Since G (c) 0; we also have
L (c; ) U (c) ; and therefore the last inequality is an equality, proving that G (c) = 0:
(2 =) 1) Condition 2 implies that U (c) = L (c; ) L (x; ) U (x) for every x
such that G (x) 0: Therefore, U (c) = J (0) ; and condition 1 follows by the last lemma.

24
Remark 38 (a) Given the inequalities G (c) 0 and 0; the restriction G (c) = 0 is
known as complimentary slackness, since it is equivalent to G (c)i < 0 =) i = 0; for
every coordinate i: Intuitively, a constraint can have a positive price only if it is binding.
(b) Condition 2 of Theorem 37 is sometimes equivalently stated as a saddle-point con-
dition: L ( ; ) is maximized over C at c; and L (c; ) is minimized over Rn+ at :

Assuming the existence of a maximum, the above global optimality conditions are
applicable if and only if @J (0) is nonempty. Convexity-based su¢ cient conditions for this
to be true are given below. Besides convexity, the key assumption is the so-called Slater
condition:

There exists x 2 C such that G (x)i < 0 for all i 2 f1; : : : ; ng : (12)

Proposition 39 Suppose that C; U and G are convex, J (0) is …nite and the Slater con-
dition (12) holds. Then @J (0) 6= ;:

Proof. One can easily verify that sub(J) is convex and has (0; J (0)) on its boundary.
By the supporting hyperplane theorem, sub(J) is supported at (0; J (0)) by some nonzero
( ; ), and therefore J ( ) implies J (0) for all 2 Rn : If < 0; then
one obtains a contradiction with = 0: The Slater condition guarantees that J ( ) > 1
for all su¢ ciently close to zero, and therefore 6= 0: The only possibility is > 0; in
which case = 2 @J (0) :

12 Local Optimality Conditions

In this section we develop local optimality conditions in terms of gradients, exploiting the
fact that small feasible perturbations near an optimum cannot improve the optimization’s
objective.
The following local optimality conditions under linear constraints are su¢ cient for most
of this text’s applications.

Proposition 40 Consider a function U : D ! R; where D X; the set fB1 ; : : : ; Bn g of

linearly independent vectors in X and the linear manifold

M = fx 2 X : (Bi j x) = bi for i = 1; : : : ; ng ; where b 2 Rn :

25
Suppose that the gradient of U at c 2 D0 \ M exists. Then
n
X
U (c) = max U (x) implies rU (c) = i Bi for some 2 Rn :
x2D\M
i=1

The converse implication is also valid if U is assumed to be concave.

Proof. Suppose U (c) = max fU (x) : x 2 D \ M g and consider any vector y in the
linear subspace
L = M c = fx 2 X : (Bi j x) = 0; i = 1; : : : ; ng :

Setting to zero the derivative at zero of U (c + y) as a function of results in the

?
orthogonality condition (rU (c) j y) = 0: Therefore, rU (c) 2 L : By Proposition 22,
L? = M ? = span(B1 ; : : : ; Bn ) and therefore rU (c) is a linear combination of the Bi :
Conversely, suppose U is concave and rU (c) 2 span(B1 ; : : : ; Bn ) : Consider any x 2 D\M;
and let y = x c 2 L: Using the gradient inequality and the fact that rU (c) is orthogonal
to y; we conclude that U (x) U (c) + (rU (c) j y) = U (c) :

Finally, we derive the Kuhn-Tucker optimality conditions for last section’s prob-
lem (P0 ) :

Theorem 41 Suppose that c 2 C 0 solves problem (P0 ) : U (c) = J (0) and G (c) 0:
Suppose also that the gradients of U and G at c exist, and that

for some h 2 X; Gi (c) + (rGi (c) j h) < 0 for all i: (13)

Then there exists some 2 Rn such that

rU (c) = rG (c) ; G (c) = 0; 0:

Proof. An exercise shows that optimality of c implies that for any h 2 X;

Gi (c) + (rGi (c) j h) < 0 for all i =) (rU (c) j h) 0:

The following two convex subsets of Rn R are therefore nonintersecting:

A = f( ; ) : i < 0 for all i; and > 0g ;

B = f( ; ) : for some h 2 X; G (c) + (rG (c) j h) and (rU (c) j h)g :

26
By the separating hyperplane theorem (Corollary 33), there exists nonzero ( ; ) 2 Rn R
that separates A and B; and therefore

0 and 0 =) + 0; (14)
[G (c) + (rG (c) j h)] + (rU (c) j h) 0 for all h 2 X: (15)

Condition (14) implies that 0 and 0: If = 0; assumption (13) is violated (why?).

After proper scaling, we can therefore assume that = 1; and the Kuhn-Tucker conditions
follow from condition (15) and the inequalities 0 and G (c) 0:

13 Exercises
1. Prove Proposition 3.

2. (a) Given the inner product space X; prove the parallelogram identity:

kx + yk2 + kx yk2 = 2 kxk2 + 2 kyk2 ; x; y 2 X:

Interpret this equation geometrically to justify its name.

(b) Verify the equations in the proof of Theorem 19.

3. Suppose D is an open convex subset of X; f : D ! R is concave, and x 2 D:

Assume that f is bounded below near x; meaning that there exist an open ball
B (x; r) D and some b 2 R such that f (y) > b for all y 2 B (x; r) : Fix any x
~ 2 D;
and show that f is bounded below near x
~ by explaining the validity of the following
steps:
(a) There exists some z 2 D and 2 (0; 1) such that x
~ = x + (1 ) z:
(b) B (~
x; r) = f y + (1 ) z : y 2 B (x; r)g :
(c) B (~
x; r) D:
(d) f is bounded below by b + (1 ) f (z) on B (~
x; r) :

4. Let X = C [0; 1] be the vector space of all continuous functions on the unit interval,
R1
with the inner product (x j y) = 0 x (t) y (t) dt:
(a) Show that the sequence fxn g de…ned by xn (t) = 1= (1 + nt) ; t 2 [0; 1] ; is Cauchy
but does not converge in X:

27
 (b) Suppose we are interested in …nding a continuous function x : [0; 1] ! R that
R1
minimizes 0 x (t)2 dt subject to the constraint x (0) = 1: Express this as a projection
problem in X; and show that the suitable projection does not exist.

5. Let X be the vector space of every real-valued sequence x = (x1 ; x2 ; : : : ) such that
P
xn = 0 for all but …nitely many values of n; with the inner product (x j y) = n xn yn :
P
Show that the functional f (x) = n nxn is linear but not continuous.

6. Give an alternative proof of the necessity of the support condition in Theorem 19(a)
that utilizes the fact that the quadratic f ( ) = kx x k2 is minimized at zero and
therefore the right derivative of f at zero must be nonnegative. Also draw a diagram
that makes it obvious that if the angle between s xS and xS x is wider than a
right angle, then there exists some 2 (0; 1) such that kx x k < kx xS k :

7. (a) Show that the vector y supports the set S at some vector s if and only if y supports
S at s. (Note that it is not assumed that s is in S, and therefore De…nition 31 is
required.)
(b) Show that the closure of a convex set is convex and conclude that, in a complete
space, projections on the closure of a convex set always exist.

8. Suppose L is a linear subspace of the …nite-dimensional vector space X: Show that

there exists a basis fB1 ; : : : ; Bd g of X such that L = span(B1 ; : : : ; Bk ) for some
k d: Using this fact, show that every linear manifold in X is closed, thus proving
the existence of projections on linear manifolds in …nite dimensions.

9. (a) State the projection expression of Proposition 22 in common matrix notation,

assuming that the underlying space is X = Rd with the Euclidean inner product
P
(x j y) = ni=1 xi yi :
(b) Prove Corollaries 23 and 24.

10. Prove Proposition 26, and derive the projection expression of Proposition 22 as a
corollary of Proposition 9.

11. Suppose P is a linear operator on the inner product space (X; ( j )) ; that is, a
function of the form P : X ! X such that P ( x + y) = P (x) + P (y) for
all ; 2 R and x; y 2 X. P is a projection operator if there exists a linear
subspace L such that P x is the projection of x on L for all x 2 X: The operator P

28
 is idempotent if P 2 = P . Finally, P is self-adjoint if (P x j y) = (x j P y) for all
x; y 2 X.
(a) Prove that P is a projection operator if and only if P is both idempotent and
self-adjoint.
(b) Apply part (a) to show that a matrix A 2 Rn n is idempotent A2 = A and
symmetric (A0 = A) if and only if there exists a full-rank matrix B such that A =
B 0 (BB 0 ) 1 B:

12. (a) Suppose X, Y and Z are vector spaces, and f : X ! Y and g : X ! Z are
linear functions such that g (x) = 0 implies f (x) = 0 for all x 2 X: Show that
L = fg (x) : x 2 Xg is a linear subspace of Z; and that there exists a linear function
h : L ! Y such that f = h g:
(b) Suppose f; g1 ; : : : ; gn are linear functionals on X such that g1 (x) = = gn (x) =
Pn
0 implies f (x) = 0; for all x 2 X: Show that f = i=1 i gi for some 2 Rn :

13. Use a projection argument to solve the problem

min x0 Qx : Ax = b; x 2 Rn 1
;

where Q 2 Rn n is positive de…nite and symmetric, A 2 Rm n is full rank and

b 2 Rm is in the range of A:

14. Consider the problem of …nding a vector x of minimum norm satisfying (x j yi ) i;

i = 1; : : : ; n; where the vectors y1 ; : : : ; yn are linearly independent. Use projection
arguments to show that the problem has a unique solution and to characterize the
solution as a system of (in)equalities.

15. Suppose that the random vector y = (y1 ; : : : ; yn )0 is generated by the model y =
A + "; where A 2 Rn m is a known matrix, 2 Rm is an unknown vector, and "
is an unobservable zero-mean random vector, valued in Rn ; with variance-covariance
matrix E [""0 ] = ; assumed to be positive de…nite. We are interested in a linear
estimator of the parameter ; that is, an estimator of the form ^ = By; for some
B 2 Rm n : The linear estimator represented by B is unbiased if BA = I; since then
E ^ = for every choice of : Using projection theory, determine the unbiased linear
estimator that minimizes the variance of ^ i i for every i 2 f1; : : : ; mg :

29
 16. (1) Suppose X = l2 , the space of square summable sequences with the inner product
P
(x j y) = 1 n=1 x (n) y (n) : Let S be the positive cone of X; that is, the set of all x 2 X
such that x (n) 0 for all n: Show that S = S and therefore S contains no interior
points. Finally, consider any s 2 S such that s (n) > 0 for all n; and show that the
only vector y such that (y j s) (y j s) for all s 2 S is the zero vector.

17. Verify the claim of display (11) ; stating the equivalence of the supergradient property
and an associated support condition.

18. Provide a geometric explanation of complementary slackness (de…ned in Remark 38),

based on a support condition relative to the set sub(J) in Rn R:

19. In the context of Section 11, suppose that U and G are continuous. Is J necessarily
continuous at an interior point of its e¤ective domain (that is, the set where J is
…nite)? Provide a proof or a counterexample.

20. Suppose that f : X ! R is concave.

(a) Show that f 0 (x; h1 + h2 ) f 0 (x; h1 ) + f 0 (x; h2 ) for every h1 ; h2 2 X for which
the above directional derivatives exist.
(b) Show that if f 0 (x; h) exists for all h 2 X; then the function f 0 (x; ) is linear.

21. Prove the opening claim of the proof of Theorem 41. Also, provide a set of convexity-
based su¢ cient conditions for the Kuhn-Tucker conditions of Theorem 41 to imply
optimality, and prove your claim.

14 Notes
Treatments of …nite-dimensional convex optimization theory include Rockafellar (1970),
Bertsekas (2003) and Boyd and Vandenberghe (2004), the latter providing an introduction
to modern computational optimization algorithms. In…nite-dimensional extensions can
be found in Luenberger (1969), Ekeland and Témam (1999) and Bonnans and Shapiro
(2000). The Hilbert-space theory outlined in this Appendix is presented and extended in
the classic text of Dunford and Schwartz (1988), where one can also …nd the notion of weak
compactness and its interplay with convexity alluded to in Section 8.

30
References
Bertsekas, D. P. (2003): Convex Analysis and Optimization. Athena Scienti…c, Belmont,
MA.

Bonnans, J. F., and A. Shapiro (2000): Perturbation Analysis of Optimization Prob-

lems. Springer-Verlag, New York.

Boyd, S., and L. Vandenberghe (2004): Convex Optimization. Cambridge University

Press, Cambridge, United Kingdom.

Dunford, N., and J. T. Schwartz (1988): Linear Operators, Part I, General Theory.
Wiley.

Ekeland, I., and R. Témam (1999): Convex Analysis and Variational Problems. SIAM,
Philadelphia, PA.

Luenberger, D. G. (1969): Optimization by Vector Space Methods. Wiley, New York.

Rockafellar, R. T. (1970): Convex Analysis. Princeton University Press, Princeton,

New Jersey.

Skiadas, C. (2009): Asset Pricing Theory. Princeton Univ. Press, Princeton, NJ.

31