Scribd
Upload
36 views

Problem13 Solutions

This document provides solutions to 3 exercises involving constrained optimization problems. Exercise 1 involves maximizing the function f(x,y)=xy subject to the constraint x^2 + y^2 ≤ 1. It is shown that the global maximum occurs at the two points (±√(2/2), ±√(2/2)) with a value of 1/2. Exercise 2 generalizes Exercise 1 by maximizing the function f(x,y)=ax+by subject to the same constraint. It is shown that the global maximum occurs at the point (a/√(a^2+b^2), b/√(a^2+b^2)) with a value of √

Uploaded by

Co londota2
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
36 views

Problem13 Solutions

This document provides solutions to 3 exercises involving constrained optimization problems. Exercise 1 involves maximizing the function f(x,y)=xy subject to the constraint x^2 + y^2 ≤ 1. It is shown that the global maximum occurs at the two points (±√(2/2), ±√(2/2)) with a value of 1/2. Exercise 2 generalizes Exercise 1 by maximizing the function f(x,y)=ax+by subject to the same constraint. It is shown that the global maximum occurs at the point (a/√(a^2+b^2), b/√(a^2+b^2)) with a value of √

Uploaded by

Co londota2
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

OQE - PROBLEM SET 13 - SOLUTIONS

Exercise 1. We look at the inequality constraint optimization problem in R2


max f (x, y) = xy
subject to x2 + y 2 ≤ 1.
(a) Define D = {(x, y) ∈ R2 : x2 + y 2 ≤ 1} and observe that D is closed and
bounded, hence compact. The function f being continuous, it achieves a global
maximum on D.
(b) We define g : R2 → R by (x, y) 7→ g(x, y) = x2 + y 2 − 1. We consequently
compute, for each (x, y) ∈ R2 , the gradient of g to be ∇g(x, y) = (2x, 2y) which
is equal to (0, 0) if and only if (x, y) = (0, 0). However, since the constraint is not
active at (0, 0), this gives us no problems.
(c) For (x, y) ∈ R2 and λ ∈ R≥0 , we define the Lagrangian function

L(x, y, λ) = f (x, y) − λg(x, y) = xy − λ(x2 + y 2 − 1)

and consequently we impose



 ∂L (x, y, λ) = y − 2λx = 0
∂x
(KKT-1)
 ∂L (x, y, λ) = x − 2λy = 0
∂y

and also

x2 + y 2 − 1 = 0 if λ > 0
(KKT-2) .
x2 + y 2 − 1 < 0 if λ = 0

(d) Solving the systems in (c), one finds the three points
(1) (0, 0);
√ √
(2) ( 2/2, 2/2);
√ √
(3) (− 2/2, − 2/2).
(e) We compute
(1) f (0, 0) = 0;
√ √
(2) f ( 2/2, 2/2) = 1/2;
√ √
(3) f (− 2/2, − 2/2) = 1/2.
√ √
The global maximum on D has thus value 1/2 and is achieved in ( 2/2, 2/2) and
√ √
(− 2/2, − 2/2).

Exercise 2. Let a, b ∈ R and consider the inequality constraint optimization


2
problem on R given by
1
2 OQE - PROBLEM SET 13 - SOLUTIONS

max / min f (x, y) = ax + by


subject to x2 + y 2 ≤ 1.
(a) Assume that a = b = 0. Then the function f is equal to the constant funciton 0
and each element of R2 is a maximum/minimum of f in R2 and therefore the same
applies to every element of D = {(x, y) ∈ R2 : x2 + y 2 ≤ 1}.
Suppose, from now on, that a 6= 0 or b 6= 0.
(b) Let g : R → R be defined by (x, y) 7→2 +y 2 − 1 and let D be as in (a). To find
2

the maxima of f , we define, for x, y ∈ R and λ ∈ R≥0 , the Lagrangian function


L(x, y, λ) = f (x, y) − λg(x, y) = ax + by − λ(x2 + y 2 − 1)
and we impose

 ∂L (x, y, λ) = a − 2λx = 0
∂x
(KKT-1)
 ∂L (x, y, λ) = b − 2λy = 0
∂y

and also

x2 + y 2 − 1 = 0 if λ > 0
(KKT-2) .
x2 + y 2 − 1 < 0 if λ = 0

Solving the systems, we get two points:


(1) (0, 0) – corresponds to λ = 0 and refers back to (a); and
√ √ √
(2) P = (a/( a2 + b2 ), b/( a2 + b2 )) with f (P ) = a2 + b2 and correspond-
√
ing to λP = a2 + b2 /2
so P is the point in D in which f achieves its global maximum. To compute the
minima of f in D, one applies the same method to −f , finding the minimum point
√ √ √
Q = (−a/( a2 + b2 ), −b/( a2 + b2 )) with f (Q) = − a2 + b2 and corresponding
√
to λQ = − a2 + b2 /2.
(c) As λP , λQ > 0, one can reduce to an EC problem and look at the functions
√
i. LP (x, y) = f (x, y) − λP g(x, y) = ax + by − a2 + b2 (x2 + y 2 − 1)/2;
√
ii. LQ (x, y) = f (x, y) − λP g(x, y) = ax + by + a2 + b2 (x2 + y 2 − 1)/2.
The associated Hessians are then
" # " √ #
−2λ P 0 − a2 + b2 0
D2 LP (x, y) = = √
0 −2λP 0 − a2 + b2
and " # "√ #
−2λQ 0 a 2 + b2 0
D2 LQ (x, y) = = √ .
0 −2λQ 0 a2 + b2
Theorem 4.3.1 yields that P is a global maximum and Q is a global minimum of f
in D.

Exercise 3. Let U = {(x, y) ∈ R2≥0 } and consider the inequality constraint


problem on U
OQE - PROBLEM SET 13 - SOLUTIONS 3

max u(x, y) = x1/2 y 1/2


subject to x2 + y 2 ≤ 400, x + y ≤ 28.
(a–b) Define g1 : R2 → R by (x, y) 7→ g1 (x, y) = x2 + y 2 − 400 and let also
g2 : R2 → R be defined by (x, y) 7→ g2 (x, y) = x + y − 28. We compute the Jacobian
associated to g = (g1 , g2 ) at any point (x, y) ∈ R2 :
" #
2x 2y
Dg(x, y) =
1 1
and note that the system

2x − 2y = 0


x2 + y 2 = 400


x + y = 28

has no solution, so CQ is satisfied whenever both g1 and g2 are active constraint


at a point (x, y). In all other cases CQ is trivially satisfied. We now observe that,
if (x, 0), (0, y) ∈ U , then f (x, 0) = f (0, y) = 0 and, since u achieves only non-
negative values and we are concerned with finding the maximum value of f , we can
restrict our investigation to the domain U ∗ = {(x, y) ∈ U : xy 6= 0}. We define, for
(x, y) ∈ U ∗ and λ, µ ∈ R≥0 , the Lagrangian function

L(x, y, λ, µ) = u(x, y) − λg1 (x, y) − µg2 (x, y)


= x1/2 y 1/2 − λ(x2 + y 2 − 400) − µ(x + y − 28)

and we impose consequently



 ∂L (x, y, λ, µ) = 1 x−1/2 y 1/2 − 2λx − µ = 0
∂x 2
(KKT-1)
 ∂L (x, y, λ, µ) = 1 x1/2 y −1/2 − 2λy − µ = 0
∂y 2

and also



 x2 + y 2 − 400 = 0 if λ > 0

x2 + y 2 − 400 < 0 if λ = 0

(KKT-2) .


 x + y − 28 = 0 if µ > 0


x + y − 28 < 0 if µ = 0

From (KKT-1) we derive


1 −1/2 1/2 1
x y − 2λx = x1/2 y −1/2 − 2λy
2 2
which is satisfied if and only if

x−1/2 y −1/2 (y − x) = 4λ(x − y).

The latter is satisfied if and only if


1
x = y or λ = − x−1/2 y −1/2 .
4
4 OQE - PROBLEM SET 13 - SOLUTIONS

However, for any choice of (x, y) ∈ U ∗ , the value − 14 x−1/2 y −1/2 is negative and
therefore we must have x = y. Thanks to (KKT-2), we then find the point (14, 14)
for which we have f (14, 14) = 14.
(c) For i ∈ {1, 2}, define Di = {(x, y) ∈ R2 : g1 (x, u)}. Then D1 is bounded and
D1 and D2 are both closed. Since U is also closed, it follows that D1 ∩ D2 ∩ U is
closed and bounded, thus compact. The function u being continuous, it achieves
both a maximum and a minimum on D1 ∩ D2 ∩ U and therefore (14, 14) is a global
maximum of f on the given domain.

You might also like