11700063

Partner A : TE 111704071

Assignment : Non-linear
Programming

Q1
. (a) No .
Because the constraints do not satisfy the quadratic programming
(b) It's not a convex set

(c) of = (2x 1 -
10X 2 , 2x2- 10X , )
(d) Neither convex nor concave function
(e) -

Xi -
xz = -
1
,
9 ,
= -
xi -
xz2, 79 ,
(
= -
2x1 ,
-

2x2)
Xi + xz = 9 ,
92 = Xi + x z2 ,
892 = (2x, , 2x2)

KKT Conditions (1) j = 1 2X , -

10X2 + 2x, 1 , -

2x , z = O

j =
22Xz -
10X , + 2X2U , -

2x2Uz = 0

2X , Mz)

'
(2)j =
1 X, 2X , -

10X2 + 2x, u , -
= 0

j =
2 X2 2X2 -
10X 1 + 2X2U , -

2x2(z) = 0

(3) -

X,
2
- x 22 + 1 = 0

X,
2
+ xz -
9 = 0

x2 + 1)
'
(4) u, -

X, -
= 0

12 Xi + xz -
9) =
0

(5) X , X2 , 0

(6) U , 12 8

(f) No.

(9) For Xi + X -1 =
0 (inner cycle) , we know that Xi + 2-9 < O and z
= 0

)=

E
2X 1 - 10X2 (2 + 100

E
= -
22X I 2u ,
u,
Solve f =
1 , 79 , + 0 792 2x2-10X1 =
-212X2
*
1 ,
7, 0
· =
4

U, >
,
o

[x=x z= 1
· (x .. x 2) =

(E) or
(-z E) - with local max value = -

cycle)
2

For Xi +X -

9 = 0 (outer we know that X, + X2 -1 > 0 and 1 = 0

,

2X, 10X2 2U2X

E
-
=

212)
of
2(2
Solve =
049 ,
+ 12892 2X2-10x1 =
212x2 +
- = 100
Uz =
6
27, 0 2 O ,

Xi X2

E xi (= )
-

( = - 3)
=

+ xz = g+(x , xz) = or with local max value = 54

 (h) False
· Because that's only happen when the objective function is concave and the

feasible region is convex .

In this problem , neither the objective function is

concave nor the feasible region is convex

xz + 7X3
2
Q2 .
f(x) = -
x, -

2x2 -
17
-

where X
,, X2 ,
X3 ER and can be either positive or negative

(a) f(x) =
( -

2x, ,
-

2xz + 2,
-

14X3)
Initialization Iteration 1
E 0 001 X1 t( 1)
=
.
=
0 5 + .
-

X ' = (0 .
5 ,
1 , 0) =
0 5 .
-

xf(x) =
1 -

1 , 0 , 0 xz = 1 + +(0)
of(x) = 1 > 3 =
I

perform iteration I X3 =
0 + t (0)
=
0

f(x) f (0 0
*
+ + xf(x)) =
.
5 -
t ,
1 ,

t)? +
*
=
(0 8 t 0 5
=

.
5 - a =
.

X 0f(x)
*

Reset = x + +
*
= (0 .
5 , 1 , 0) + 0 5 .
( -

1 ,
0 , 0)
X' =
(0 ,
1 , 0)
Jf(x)) =
(0 ,
0 , 0
of(x) =
0 > 3 Stop the iteration

'=
(b) Yes, X 10 , 1 , 0) is the global maximum ,
as verified by the unconstraint

non-linear programming has concave objective function and Jf(x') = (0 .

0 , 0)

3
Q3 - Max X + 2x -
2x = - 02314
I S
-

(a) Given range 0 ,

4

let X = 0 Midpoint X' =

X + =
2

x =
4 2

3x7+ 4X x3
,x
= 2 -
-

= 12 + c -
8 -

8 · Since f'(2) = O ,
we reset by x'

Thus is 0
= -
2 ,
the new range , 2
 (b) Based on Newton Method
,

Xi + 1
=
Xi -

f'(Xi)
f"(Xi)

Start from Xo =
2 ,

X
, =
Xo -
f'(Xo)
f" ( Xo)
= 2 -
( -

2)
( -

4)
= 1 .
5

Q4 .
Max f(x) =
20X, -
20X , + 50x2 -
50X22 + 18X , X2

S +
.
X, + xz = 6

X, + 4X2 = 18

Xi ,
X 2 , 0

(a) KKT Conditions

(1) j 20 40X , 18Xz 6)
!
1 M Mz = O
(4)
=
+
X, xz
- -
-
,
,
+ - =
0

j 250 100X2 + 18X , Mi 41z = O

18)
= -

12 0
-

4Xz
-

x1 + -
=

(2) j =
1 X, (20 -
40X , + 18Xz -

M -
Mz) =
0 (5) X , X2 > 0

j = 2 x2 (50 -

100X2 + 18X , -
M -
41z) =
0 (6) u , +z >, 0

(3) X, + Xz -
6 = 0

X, + 4Xz -

18 = 8

40X , + 18xz -
M ,
-
12 + Y
, =
=
20

18X1 -
100X2 -
11 -

42 + Yz =

= 58

Xi + X2 + VI = 6

Xi + 4X2 + Vz = 18

Multiply the equations by (-1) to Obtain non-negative Rus

40X , -

18Xz + 11 + 1z -

,
y = 20

18X1 + 100xz + 1, + 442 -

Y2 = 58

X, + X2 + Vi - 6

X
1 + 4X2 + V2 = 10
 The modified simplex method

Min z = ,
+ 2

S t 40x , 18Xz 1, 1z + z 20
+ + ,
y [
-
-

,
.

18X1 + 100xz + 1, + 442 -

Y2 + zz = 58

X, + X2 + Vi - 6

X
1 + 4X2 + Vz = 18

Xi, X2, i, ,
2 ,, 2
, V , 2, O

X, y + Xzz + ,V, + MzVz = 0 Additional complementarity constraint

(b) Converting from min z to max-Z

min z = z ..
+ zz

=
(20 -
40x1 + 10X2 -

M, -
Mz + y, ) + (50 -
100x2 + 18X , -
M ,
-

41z +
yz)
max (-7) =
-

20 + 40X ,
-
18Xz + 1 , + 1z -

y, -
50 + 100x2 -
18X1 + 1 , + 41z -

yz
-

z -

( -

20 + 40X ,
-
18x2 + 1 . + 1z -

y, ) -

( -
50 + 100x 2 -
18X1 + 1 , + 41z -

yz)
-
z + 20 -
22X1 -

82Xz -

21 ,
-

51z + y ,
+ yz = 0

z -

22X, -
82Xz -

211 -

51z + y1 + yz = -
70

Iteration O

Coefficient of
Basic Variable RHS
= X, x zu , > z , zzy , yz v, vz

Z O -

1 -

22 -
82 -

2 -

500 /100 -
70

El I O 40-18 I 110
-
1000 20

=
2
2 0-18100 I 4 0 1 0 -

10 O 50

Vi 3 0 1100 0 0 0 0 1 0 6

V2 4 O 140000000 18

Entering Variable =
X2

Leaving Variable =
2

f(x)
Cf.
Q5 .
Max =
20X 1 + 10x2 (1)
s + X,
2
+ xz = 2Xj

=
.

u )
*

X, + 2X1 2 (2) X , (of -

=
0
GXj
X, X2 ,
O
(3)9i(x ) bi = 0 *
-

(4)
ui(9i(X ) bi
*
-
= 0

*
(5) Xj >
, 0

(6) i , 0
 = 0
(1) j = 1 : 20 -
2 x, ,
-
1z

X 212 = 0
j =
2
: 10 -
11 , .
-

(2)j = 1 : x, (20 -
2 M, X -
Mz) = 0 (5) X, 3 0 ,
Xz 8

j =
2 : xz (10 -
24 , X -

2Mz) = 0 (6) u , >, 0 , 12 0

(3) X,
2
+ xz -
1 = O

Xi + 2x2 -

2 = 0

(4)u (x , ,
2
+ xz -
1) = 0

Uz(X : + 2xz -

2) =
0

· The solution (x .. x2 , ) =
(5) satisfies all conditions .

(E) optimal
*
Since this is a convex programming problem , x =
is

Q6 . (a) · In (X . + 1) is concave function for X. 0 ·

f(x) is the sum of In (X . + 1)
Since an(X, + ) and X2 Hence f(x)
,
is a
= 0
,
= -
.

d X,
2
(x + 12 concave function for X ., 0

·
X2 is concave function for X2 , 0 and X27, 0

function
x
Since Xz is a linear and =
00

(b) The set

feasible region is convex

...
2X , + x 2 - 3

I
XI O 3
X2 3 J
3
X , X 2, 0
-

(C) KKT Conditions

1
21 =0
(1)j
-

,
=
1 :
(x , + 1)
j =
2 : 1 -

M ,
=8

I
2u 0
(2) j Xi =
=
1
-

: ,

(X , + 1)

1 2 Xz(1 M ) 0
= : -
=
.

(3)2X1 + xz -

3 = 8

(4) u , (2X 1
+ xz -

3) = 0

(5) X 1 >, 0 , X 2, 0

(6) 1, , 0
(d) Find all optimal solution

Steps : .
1
U
, > 1, from condition 1 (j =
2)
X
, 0 , from condition 5

.
2 Therefore,
x + 2 ,
< o

.
3 Therefore
, X, =
0 from condition 2(j =
1)
4
. U , #0 implies that 2X . + X2-3 =
0 from condition 4

.
5 From steps 3 & 4 , we can imply that X2 =
3

.
6 X2 0
, implies that ,
= 1 from condition 2(j =
2)
.
7
The solution (X , X2
, ) =
(0 , 3 , 1) satisfies all the conditions .

*
Since this is a convex programming problem
, X =
(0 3)
,
is optimal