13 .

2-10
po!nts E , 0) and 1-12 0
for !nstance take ,

m!n 2-x,* + ex22

S to we!ght 112 +
& (2 0) ,
+
& (12 ,
%) =
100
x( x2x2 x! x2
.

+ 2
+ + =
Feas .
Sol , !s not a convexset .

r2

b) -
X, 2 - x22-2 + we are
look!ng at the feas, reg!on .

-
22 12 XI
↓ ↓
., (X1)
9 92(X2) Feas!ble need
2 reg ,
not to be
09(x 1 )

2ge(x) y both
s
convex
Ver!fy !t also
-
12 -
> .
set
by
-X , 2 and we!ghts
cons!der!ng 2
po!nts
as we d!d !n parto .

13 3 y =
(1 y2)
4
,
-

Y2
e41
.

a) Let X ,
=
,
X = e
2 "
f(x) 2
-

2x!
-

=
x2 + xz 29 -

Y2 292
f(2)
-
-

↓ m!n
=
2e + e

X = X 1, X2 S to
2y1 + 232
.

y1 + 32
44x2 + X12x22x12 42 +
C = 12

X 1 1X2 >, 0
e
Y
,
e0 , for E (31 32)
,

-32-3-232
-23

b)
,

& , 01 for Fly2) [1]

cond!t!ons hold
e ye-24230
241 Y2
for (2) [2] s!nce all
-
-

2
= +
that the
we can
say
funct!on that we cons!der
2y1
-

32 A convex

!
.
-

4, 0 [3] ↑
[1] [2] [3] [4] 7, 0
[332
, ,
,

& (1)x(2) -
= 30V[4]
 13 3 .
-

f(x) 10x1 + 20x2 10 10t

max
a) 1031 + 2032
+
max
=
+

3) x1 + 4x2 + 20 S .
to
S to
y1
.

+ 342[50t
X! + 3x2 50 33 + 232 [80t

3x1 + 2x2 80
In . Y2 t > 10
X1 , +270 1
331 + 44 + 207 = 1
t -
=

3x1 + 4x2 + 20
we constra!ned related w!th

13 4 -

3 Y1 = X1 t .

t =
.

3x1
y2 =
X2t + 4x2 +20

f(x) =
48x5 + 42x3 -

3 . 5X -

16x6 61x4 -

16 5 .
X2

9) B!sect!on method 2 = 0 .
08

f!nd !n!t!al andX

In!t!al!zat!on : 1 by !nspect!on.

Select an !n!t!al solut!on

=
Iterat!ons select a new X =
1.
Evaluate at x = stop rule : If -* Ce 10 16) .

reset x !terat!on
If
=

.
2
20

Y =
X O -
-

1 4312 -

16 69 .

If o
11
3 .
1 -100 -

1 1 5 .
0 . 25 0 .
3047

2 0
+

.
153 0 25 .
1 .
5 0 .
835 0 .
2482

3 -
0 .
923 0 . 21 0 .
875 0 5625 .
0 .
3125
0 01 0 5625 0 4063 0 3124
4 0 21
-

.
. . .
.

0 4844
:

5 0 09 .
0 40640 5625
.
.
.

0 312
.

= 0 15620 16 ~
. 5625
0 . 4064 = 0 .
.

Stop
b) Newton X1 =
0 . 4844
2 = 0 .
00 X! = 1
Iterat!on :

. Calculate
1 f'(x) and f"(x)

. Set
2 X!t1 =

X!-!l
.
3 Stop rule W!tl-X!) e X!t1 !s the opt!mal solut!on
(x!H-X!1 <t reset ! !t and perform another !terat!on .
i
 f(x) + '(x) +"(x) X! + )(x + 1
-
X!)
!terat!on : X!

I 1 O -

23 .
5 -

93 0 747 .
0 .
263

0 747 0 305 8 496 36 04 0 512 0 286

2
-
-
. .
.
.
. .

3 . 512
0 0 312-2 677
. .
-
15 703 .
0 .
341 0 17 .

3410 312 0 768 7 588 0

24 101
4 0 0
- -
. . .
. .
.

303 091 6 461 226

5 0 24 0 -0 0 0 014
-
.
.
. .
. .

6 0 226
. 0 3 .
0 .
01
-

6 68
.
0 23 0 . 0002 Le
.

s
-

2
Us!ng Hess!an Matr!x

Of f!x the matr!x all pr!nc!pal

- >
-
!jth entry
.

=
n n j
Ax!Xj
- -

The Grad!ent Search procedure

* NP - W!nston >
-
pablo
X x +
t * f(x)
=

"Genet!c Algor!thm" concave *

f(x) + = v f(x)
AND max
II
"Warehouses t70

The grad!ent search pro .

example X! =
xj +
t()x =
x

Cont!ne unt!l f(x) = 0

i
 13 5 - 1 .

# summary Grad!ent Search

of the

In!t!al!zat!on : select t and tr!al solut!on x

Go f!rst to the
stopp!ng rule .

Iterat!on :

1) Express f(x' + +Vf(x))) as a funct!on of + by sett!ng

((x
xj xj = +
1
forj = . 2 ..
n

= y
then substut!de express!ons !nto F(x)

var!able
2) use a search procedure for unconst .

Opt! .
to f!nd =t
*
that max f!x' + tufx'llover

3) Reset X = x + uf(x) then go the stoprule .

S
Evaluate Vf(x) at least

grad!ent >
- f!nd d!rect!ons for both var ables X= X)

!f (fr
.

al Check

j = 1 , 2 .. n

Iterat!on
X' V f(x))
*
x( + z!f(x))t x+ ++ f(x)

f(x) =
2X , X2 -
x2
-

X! -

2x22

1 (1 , 1) (01 -

1) (1 e-t) , 0 .
25 ( ,
0 .
75)
Yf(x) =
2x2 -

2x) ,
(X) +1
-

4x2)
↓
(10 75) 75) Vf(x) (01-1)
2 (05 0) 5toes) 0 5 10 75 for the 1st !terat!on
0 1
0 .

,
.
. .
.

* + twf(y') for the 1st !t :

3 10 75 1 .
0 .
75)
(1 , 1) + +(0 -
1) = ( , 1 - t)

4 f (1 , 1 -
t) =
2X(X(1 t) -

+ (1 -

t) -
12 -
2(1 -
t)2
= t -

2t2
(t -

2 +2)) =
1 -

4t = 0

Ff(x)) =

(2x2 -

2x) ,
2x, + 1
-

4x2) t = 0 . 25
X1 =
(1 ,
0 .
75)

= f(x)) =
(2x075 -
2x1) = - 0 .
5 ,
2 +1
-

4x 05 = d =
70 5 .

,
d
x + + yf(x) (1 = ,
0 . 75) + 1 -
0 .
5t ,
d) =

((1 -
0 . 5) 1 0 .
75)
f (1 0 5t , 0 75)
-

.
.
=
2 (1 -
0 5+)(0
.
.

75) + 0 .
75 -
11-0 5772 .
-
210 . 7572
↓
75) 1
F (1-0 57 .
, 0 . 75) = 0 75.

A
= 1-0 57 .
X + t yf(x)) =

(1 ,
0 .
+

t = 0 5 .

-
i
 b 2x1 0 x 42
=

= 2x2 -
=

0
1 4x2 =

=
4
0 2x1 + -

2x +
=
-

X1 =
X2 = 0 . 5
-
soln
Opt .

search
0 625 grad!ent
10 75 .
opt Soln
,
.

near
method g!ves
.

f(x) 2 10 5) 2 0 5-10 512 210 512 25

-
+ = 0
.
= . -

!
.
.
. .

0 24-
He
f(x) = 2 . (075)10 625) .
+ 0 .
625 -

10 .
751 -

210 62512 .
= .

c) Tr!al seln
↓2

1
-
8 . . . . . . .

(1 , 1) 0 75 o-
same log c after last soln

o↑
.

(1 , 0 .
75)
10 75
!
.

,
0 .
75)
625)
10 .
75 , 0 .

.
0

0 5 .
&

d!s !
13 .
5 -
3

( ! X2) =
(0 , d !n!t .
sel w!th t = 0 3
.

f(x) =
8x1 -
X! -
12x2 -
2x22 + 2x , x2

Y f(x) =
( -

2x + 2x2 1
-
12 -

4x2 + 2x1)

X2 X1 4
0
-
- =

Herat!on
XVf(x))x f(x) + Ax + E-(x)
8 2x +
2x2 =
-

+
X 2x2 = 6
0
-

12 -

4x2 + 2x) =

X2 2 X2 2

&
-
= = -

X1 = 2

0
1 (1 529 1 29) 2X-4 =

(2 -2) Opt seln

-

0 . 0 8 .
-12 8t .
-12t 0 .
191 .
-

2 .
, .

X2 = -

x1 = 2

2 1529 , 2.29 0 361

.
, 0 219 .

1 529 + 0 36t
.
·

-22 1 .

31 2 .
002 2

3 2 002 . ,
-

Zen (0 ,
0 .
003)
near opt both < E
obta ned
bygraerch STOP !
i
i
i
 13 .

5-4

!terat!on x -f(x) X + twf(x) -Ax + +* v+(x)

↑
k near
412 4t It
100 ,

3
Y f(x) =
(4 + 2x1 -

4X , -

2x212 2x
-
-

4x2)
3
f'(t) 20 52t 3X256t
2672 25674
- -

↑ (4t , 2t)
=

= 20t -

B!sect!on method
5
*
f!nd + >
- apply 2 !terat!ons of .

t = o
,
E = +

nowt
It . E f(t)

0 -

01 = 05 -

144

1 -

134/8 & 0 5 .
0 .
25 -

14

↓1
-
940 0 0 .
25 0 .
125

t =
0 . 5 1st !t

t = 0 .
25 2nd !t