Scribd
Upload
8 views

Note 1

Uploaded by

chaymae
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
8 views

Note 1

Uploaded by

chaymae
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

1 Convex Sets and Functions

1.1 Convex Sets


Definition:(Convex sets) A subset C of Rn is called convex if

λx + (1 − λ)y ∈ C, ∀ x, y ∈ C, ∀λ ∈ [0, 1].

Geometrically, it just means that the line segment joining any two points in a
convex set C lies in C.

Figure 1: Examples of convex and non-convex set

Definition:(Convex
Pm combination)
Pm Given x1 , ..., xm ∈ Rn , an element in the
form x = i=1 λi xi , where i=1 λi = 1 and λi ≥ 0 is called a convex combina-
tion of x1 , ..., xm .

Proposition: A subset C of Rn is convex if and only if contains all convex


combination of its element.

Proof. Suppose C isP convex. We will show by induction that it contains all
m
convex combination i=1 λi xi of its elements.
The case m = 1, 2 is trivial, so suppose all convex combination of k elements
lies in C, where k ≤ m. Consider
m+1
X m+1
X
x := λi xi , where λi = 1
i=1 i=1

If λm+1 = 1, then λ1 = · = λm = 0. Then x ∈ C. So assume λm+1 < 1, then


m m
X X λi
λi = 1 − λm+1 and =1
i=1 i=1
1 − λm+1

1
Pm λi
Then y = i=1 1−λm+1 xi ∈ C. Hence

x = (1 − λm+1 )y + λm+1 xm+1 ∈ C

The other direction is trivial.

Proposition: Let C1 be a convex set of Rn and let C2 be a convex set pf Rm .


Then the Cartesian product C1 × C2 is a convex subset of Rn × Rm .

1.1.1 Examples of Convex Sets


(a) Open and closed balls in Rn .
(b) Hyperplanes: {x : ha, xi = b, a ∈ Rn , b ∈ R}.
(c) Halfspaces: {x : ha, xi ≤ b, a ∈ Rn , b ∈ R}.
(d) Non-Negative Orthant: Rn+ = {x ∈ Rn : x ≥ 0}.
(e) Convex cones: C is called a cone if αx ∈ C, ∀α > 0, x ∈ C. A cone which
is convex is called a convex cone.

Proposition: Let {Ci | i ∈ I} be a collection of convex sets. Then:


(a) ∩i∈I Ci is convex, where each Ci is convex.
(b) C1 + C2 = {x + y | x ∈ C1 , y ∈ C2 } is convex.
(c) λC is convex for any convex sets C and scalar λ. Furthermore, (λ1 +λ2 )C =
λ1 C + λ2 C for positive λ1 , λ2 .
(d) C o , C are convex, i.e. the interior and closure of a convex set are convex.
(e) T (C), T −1 (C) are convex, where T is a linear map.

Proof. Parts (a)-(c), (e) follows from the definition (Exercise!). Let’s prove (d).
Interior Let x, y ∈ C o . Then there exists r such that balls with radius r centred
at x and y are both inside C.
Suppose λ ∈ [0, 1] and ||z|| < r. By convexity of C, we have,

λx + (1 − λ)y + z = λ(x + z) + (1 − λ)(y + z) ∈ C

Therefore, λx + (1 − λ)y ∈ C o . Hence C o is convex.


Closure Let x, y ∈ C. Then there exists sequences {xk } ⊂ C, {yk } ⊂ C such
that xk → x, yk → y. Suppose α ∈ [0, 1]. Then for each k,

λxk + (1 − λ)yk ∈ C

But λxk + (1 − λ)yk → λx + (1 − λ)y ∈ C. Hence, C is convex.

2
1.2 Convex and Affine Hulls
1.2.1 Convex Hull
Definition:(Convex Hull)
Let X be a subset of Rn . The convex hull of X is defined by
\
conv(X) := {C| C is convex and X ⊆ C}

In other words, conv(X) is the smallest convex set containing X.


The next proposition provides a good representation for elements in the convex
hull.

Proposition: For any subset X of Rn ,


m m
X X
conv(X) = λ i xi | λi = 1, λi ≥ 0, xi ∈ X
i=1 i=1
 Pm Pm
Proof. Let C = i=1 λi xi | i=1 λi = 1, λi ≥ 0, xi ∈ X . Clearly, X ⊆ C.
Next, we P check that C is P convex.
p q
Let a = i=1P P j=1 βj bj be elements of C, where ai , bi ∈ C with
αi ai , b =
αi , βj ≥ 0 and αi = βj = 1. Suppose λ ∈ [0, 1], then
p
X q
X
λa + (1 − λ)b = λαi ai + (1 − λ)βj bj .
i=1 j=1

Since
p
X q
X p
X q
X
λαi + (1 − λ)βj = λ αi + (1 − λ) βj = 1
i=1 j=1 i=1 j=1

we have λa + (1 − λ)b ∈ C. Hence, C is convex. Also, conv(X) ⊆ C by the


definition of conv(X).
P
Suppose a = λi ai ∈ C. Then since each ai ∈ X ⊆ conv(X) and conv(X) is
convex, we have a ∈ conv(X). Therefore, conv(X) =C.

You might also like