Scribd
Upload
43 views4 pages

Algorithms For Optimal Decisions Tutorial 1 Answers

This document contains solutions to exercises on convexity from a tutorial on algorithms for optimal decisions. The solutions show that: 1) The intersection of any number of convex sets is convex. 2) The sum of two convex functions is convex. 3) The set of points where a convex function is below a bound is convex. 4) The constraints and objective function of a nonlinear optimization problem are convex.

Uploaded by

Sofyan Dc
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
43 views4 pages

Algorithms For Optimal Decisions Tutorial 1 Answers

This document contains solutions to exercises on convexity from a tutorial on algorithms for optimal decisions. The solutions show that: 1) The intersection of any number of convex sets is convex. 2) The sum of two convex functions is convex. 3) The set of points where a convex function is below a bound is convex. 4) The constraints and objective function of a nonlinear optimization problem are convex.

Uploaded by

Sofyan Dc
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/ 4

Algorithms for Optimal Decisions

Tutorial 1
Answers

Exercise 1 Show that the intersection S of any numbers of convex sets Si


is a convex set.
Solution : Take any two elements x1 , x2 from the intersection set S = i Si .
In order to prove that the intersection S is convex we need to prove
that
x1 + (1 )x2 S, [0, 1].
(1)
Since x1 , x2 S it follows that x1 , x2 Si ,

i.

Because each Si is a convex set then, for all i:


x1 + (1 )x2 Si ,

[0, 1].

(2)

Hence the point x1 +(1)x2 belongs to all the sets Si for [0, 1].
Consequently it belongs to the intersection S = i Si of all these sets.
Therefore we proved that for any two elements x1 , x2 in the intersection
S = i Si and for any [0, 1] the following holds:
x1 + (1 )x2 S,

[0, 1].

(3)

According to the definition of a convex set, the set S = i Si is also a


convex set.

Exercise 2 Show that if f (x) and g(x) are convex functions on a convex set
S, then their sum
h(x) = f (x) + g(x)
(4)
is also a convex function on S.
Solution : Take any two elements x1 , x2 from the set S. To prove that the
sum f (x) + g(x) is a convex function we need to show that:
f (x1 + (1 )x2 ) + g(x1 + (1 )x2 )
(f (x1 ) + g(x1 )) + (1 )(f (x2 ) + g(x2 )).

(5)

Since f and g are convex functions we have that for any two points
x1 , x2 S, and [0, 1] the following holds:
f (x1 + (1 )x2 ) f (x1 ) + (1 )f (x2 )
g(x1 + (1 )x2 ) g(x1) + (1 )g(x2 ).

(6)
(7)

Adding (6) and (7) we have


f (x1 + (1 )x2 ) + g(x1 + (1 )x2 )
f (x1 ) + (1 )f (x2 ) + g(x1 ) + (1 )g(x2 ) =
= (f (x1 ) + g(x1 )) + (1 )(f (x2 ) + g(x2 )),
which shows that (5) holds and consequently the sum f (x) + g(x) is a
convex function.
Exercise 3 Show that if f (x) is a convex function, then the set
L = {x Rn

| f (x) b}

(8)

is a convex set.
Solution : We need to prove that for every x1 , x2 L the point x1 + (1
)x2 is also in L.
For any two elements x1 , x2 L we have:
f (x1 ) b, f (x2 ) b
f (x1 ) b, (1 )f (x2 ) (1 )b.
2

Adding the above two inequalities we have:


f (x1 ) + (1 )f (x2 ) b + (1 )b = b.

(9)

Since the function f (x) is convex we also have:


f (x1 + (1 )x2 ) f (x1 ) + (1 )f (x2 ).

(10)

From (9) and (10) it follows that


f (x1 + (1 )x2 ) f (x1 ) + (1 )f (x2 ) b,

(11)

which shows that the point x1 + (1 )x2 L and consequently the


set L is convex.
Exercise 4 Consider the nonlinear problem:
min

f (x) = x21 + x22 4x1 + 4

s.t.

g1 (x)
g2 (x)
g3 (x)
g4 (x)

=
=
=
=

x1 x2 + 2 0
x21 + x2 1 0
x1 0
x2 0.

(12)

1. Show that the constraints define a convex set;


2. Show that the objective function f (x) is convex.
Solution :
(a) The feasible region (i.e. the set defined by the constraints of the problem) is convex because:
(i) constraints g1 (x), g3 (x) and g4 (x) are linear and hence concave.
(Remember that a linear function can be both concave and convex.)
(ii) constraint g2 (x) is nonlinear. To check whether it is concave or
not we need to find its Hessian matrix:

H2 =

2 g2
x21
2 g2
x2 x1

2 g2
x1 x2
2 g2
x22

"

2 0
0 0

(13)

and show that it is negative semidefinite, i.e.


v R2 ,

v t H2 v 0.

(14)

The matrix H2 is negative semidefinite because for every vector


v t = (v1 , v2 ) R2 we have:
t

v H2 v = (v1 , v2 )

"

2 0
0 0

v1
v2

= 2v12 0.

(15)

Therefore all the functions gi , i = 1, 2, 3, 4 which define the feasible


region are concave functions. We have concave functions, and from
the previous example, sets:
Li = {x Rn

|gi (x) 0},

i = 1, 2, 3, 4

(16)

are convex (show it!). Feasible region F = i Li is an intersection


of convex sets, therefore also convex.
(b) To show that the objective function f (x) is convex we need to show
that its Hessian matrix

H =

2f
x21
2f
x2 x1

2f
x1 x2
2f
x22

"

2 0
0 2

(17)

is positive semidefinite.
Matrix H is positive semidefinite because for any v t = (v1 , v2 ) R2
we have:
t

v Hv = (v1 , v2 )

"

2 0
0 2

v1
v2

= 2v12 + 2v22 0.

(18)

You might also like