Problem 10 Problem Set

10.3A Page 414

Maximize z = y1y2…yn
subject to y1+y2+…+yn = c,
yi  0
Thus there are n stages to this problem. At
stage i, we have to choose the variable yi.
The state of the problem at stage i is defined
by the variable xi, which represents the sum
of the variables to be decided at stages i,
i+1, …, n. Thus
x1  y1  y2  ...  yn
x2  y2  y3  ...  yn
.
.
xn  y n
Let Fi ( xi ) be the optimal return for

stages i, i+1, …, n.
Stage n: Thus xn = yn
Optimal return for this stage =

Fn ( xn )  max{ yn }  xn
yn
Stage n-1:
Optimal return for this stage =

Fn1 ( xn1 )  max{ yn1  Fn ( xn )}

yn1

 max{ yn1  Fn ( xn1  yn1 )}

yn1

2
 max{ yn1  ( xn1  yn1 )}   xn 1 
yn1  2 
 
occurs when yn-1 = (xn-1)/2
 Stage n-2:
Optimal return for this stage =
Fn  2 ( xn  2 )
 max{ yn  2  Fn 1 ( xn  2  yn  2 )}
yn  2

( xn 2  yn 2 )
3
 xn 2 
2
 max{ yn  2  } 
yn  2 4  3 
occurs when yn-2 = (xn-2)/3
Stage i:
Optimal return for this stage (by
induction on i) =
n i 1
 xi 
Fi ( xi )   
 n  i 1 

occurs when yi = (xi)/(n-i+1)

Stage 1: Optimal return for this stage =

n n
 x1   c 
F1 ( x1 )      
 n  n
occurs when y1 = x1/ n = c/ n
y2 = x2/(n-1) = (x1-y1)/(n-1)
= (c- c/n)/(n-1) = c/n

Similarly, yi = c/n for all i.

Problem 12 Problem Set
10.3A Page 415
Maximize z = (y1+2)2+y2y3+(y4-5)2
subject to y1+y2+y3+y4  5,
yi  0, integers
In order to get a proper decomposition of
the objective function , we rewrite it as

Maximize z = (y1+2)2+(y2-5)2 +y3y4

subject to y1+y2+y3+y4  5,
yi  0, integers

Thus there are 4 stages to this problem. At

stage i, we have to choose the variable yi. The
state of the problem at stage i is defined by the
variable xi, which represents the sum of the
variables to be decided at stages i, i+1, …, 4.
Thus
x1  y1  y2  y3  y4
x2  y2  y3  y4
x3  y3  y4
x4  y 4

Let Fi(xi) be the optimal return for stages

i, i+1, …, 4.
n=4 Thus x4 = y4 = 0,1,2,3,4
Optimal return for this stage =
F4 ( x4 )  max{ y4 }
y4
x4 F4(x4) y 4*
0 0 0
1 1 1
2 2 2
3 3 3
4 4 4
5 5 5
 n=3 Here
F3 ( x3 )  max{ y3 * F4 ( x3  y3 )}
y3

y3 y3*F4(x3-y3) F3(x3) y 3*
x3 0 1 2 3 4 5
0 0 - - - - - 0 0
1 0 0 - - - - 0 0,1
2 0 1 0 - - - 1 1
3 0 2 2 0 - - 2 1,2
4 0 3 4 3 0 - 4 2
5 0 4 6 6 4 0 6 2,3
n=2 Here
F2 ( x2 )  max{( y2  5)  F3 ( x2  y2 )} 2
y2

y2 ( y2 y5) F
*f2 (x -y3 ()x2  y2 )
3 4 3 3 F2(x2) y 2*
x2 0 1 2 3 4 5
0 25 - - - - - 25 0
1 25 16 - - - - 25 0
2 26 16 9 - - - 26 0
3 27 17 9 4 - - 27 0
4 29 18 10 4 1 - 29 0
5 31 20 11 5 1 0 31 0
n=1 Here
F1 ( x1 )  max{( y1  2)  F2 ( x1  y1 )} 2
y1

y1 ( y1 y2) 2
 -y
*f (x F2 () x1  y1 )
3 4 3 3 F1(x1) y 1*
x1 0 1 2 3 4 5
0 29 - - - - - 29 0
1 29 34 - - - - 34 1
2 30 34 41 - - - 41 2
3 31 35 41 50 - - 50 3
4 33 36 42 50 61 - 61 4
5 35 38 43 51 61 74 74 5
Thus the optimal solution is:

y1  5, y2  0, y3  0, y4  0
And the optimal value is z = 74
Hillier and Lieberman Problem 11.3-13
Page 573
Maximize z  y1 y y
2
2
3
3

Subject to
y1  2 y2  3 y3  10
y1  1, y2  1, y3  1,

and y1 , y2 , y3 are integers.

Solution: There are 3 stages; at stage i, we
decide yi. At stage i, the problem will be in
state xi, that represents the " amount " left
for allocation.
If Fi(xi) is the optimum return for stages
i,i+1, …,3, then we have the following
recurrence relations:
F3 ( x3 )  x33

F2 ( x2 )  max y22  F3 ( x2  2 y2 )
y2

F1 ( x1 )  max  y1  F2 ( x1  y1 )
y1
Stage 3
Here y3 = 1, 2, or 3.

x3 F3(x3) y 3*

0, 1, 2 0 0
3, 4, 5 1 1
6, 7, 8 8 2
9, 10 27 3
 F (
Stage 2 2 2x )  max y
y2

2  F3 ( x2  2 y2 )
2

y2 y22  F3 ( x2  2 y2 )
x2 1 2 3 4 5 F2(x2) y2*
5 1 - - - - 1 1
6 1 - - - - 1 1
7 1 4 - - - 4 2
8 8 4 - - - 8 1
9 8 4 9 - - 9 3
10 8 32 9 - - 32 2
 Stage 1 F1 ( x1 )  max  y1  F2 ( x1  y1 )
y1

y1 y1  F2 ( x1  y1 )
x1 1 2 3 4 5 F1(x1) y1*
10 9 16 12 4 5 16 2

Thus the optimal solution is:

y1 = 2, y2 = 1, y3 = 2 and optimal z = 16.
 Brute-force Verification
SNo y1 y2 y3 y1+2y2+3y3 y1*y22*y33
1 1 1 1 6 1
2 1 1 2 9 8
3 1 2 1 8 4
4 1 3 1 10 9
5 2 1 1 7 2
6 2 1 2 10 16
7 2 2 1 9 8
8 3 1 1 8 3
9 3 2 1 10 12
SNo y1 y2 y3 y1+2y2+3y3 y1*y22*y33
10 4 1 1 9 4
11 5 1 1 10 5
Hillier and Lieberman Problem 11.3-10
Pages 572- 573

Consider an electronic system consisting of

four components, each of which must work
for the system to function. The reliability of
the system can be improved by installing
several parallel units in one or more of the
components. The following table gives the
probability that the respective components
will function if they consist of one, two, or
three parallel units:
 Probability of functioning
Component
Parallel units 1 2 3 4
1 0.5 0.6 0.7 0.5
2 0.6 0.7 0.8 0.7
3 0.8 0.8 0.9 0.9

The probability that the system will function is

the product of the probabilities that the
respective components will function.
 The cost (in hundreds of dollars) of installing
one, two, or three parallel units in the respective
components is given by the following table:

Cost
Component
Parallel units 1 2 3 4
1 1 2 1 2
2 2 4 3 3
3 3 5 4 4
 Because of budget limitations, a maximum of
$1,000 can be expended.
Use Dynamic Programming to determine
how many parallel units should be installed in
each of the four components to maximize the
probability that the system will function.
Solution: There are 4 stages. At stage i, we
decide the number yi of component i to be
installed. At stage i, the system is in state xi,
namely the amount to be allocated to the
stages i, i+1, …,4. All amount is in hundreds
of dollars.
Our objective is to maximize the reliability of
the system, namely,
z  p1 ( y1 )  p2 ( y2 )  ...  pn ( yn )
where pi(yi) = probability of component i
functioning when yi number of component i
are installed (and is given in the first table)
If Fi(xi) is the optimal reliability of the system
for stages i,i+1, …, 4 we then have the
recursive relation
Fi ( xi )  max{ pi ( yi )  Fi 1 ( xi  ci ( yi ))}
yi
i=1,2,3,4
where pi(yi) = probability of component i
functioning when yi number of component i
are installed (and is given in the first table)
and ci(yi) is the cost of yi number of
component i (and is given in the second
table).
 Stage 4
x4 F4(x4)= p4(y4) y4*

2 0.5 1
3 0.7 2

 4,  6 0.9 3
 Stage 3
y3 p3(y3) * F4(x3-c3(y3))
x3 1 2 3 F3(x3) y3*
0.7x0.5
3 = 0.35 - - 0.35 1
0.7x0.7
4 = 0.49 - - 0.49 1
0.7x0.9 0.8x0.5
5 = 0.63 = 0.40 - 0.63 1
0.7x0.9 0.8x0.7 0.9x0.5
6 = 0.63 = 0.56 = 0.45 0.63 1
0.7x0.9 0.8x0.9 0.9x0.7
7 = 0.63 = 0.72 = 0.63 0.72 2
 Stage 3 (Continued)
y3 p3(y3) * F4(x3-c3(y3))
x3 1 2 3 F3(x3) y3*
0.7x0.9 0.8x0.9 0.9x0.9
8 = 0.63 = 0.72 = 0.81 0.81 3
0.7x0.9 0.8x0.9 0.9x0.9
9 = 0.63 = 0.72 = 0.81 0.81 3
0.7x0.9 0.8x0.9 0.9x0.9
10 = 0.63 = 0.72 = 0.81 0.81 3

(Note: For stage 3, min x3 = 3, max x3 = 7.)

 Stage 2
y2 p2(y2) * F3(x2-c2(y2))
x2 1 2 3 F2(x2)
y 2*
0.6x0.35
5 = 0.210 - - 0.210 1
0.6x0.49
6 = 0.294 - - 0.294 1
0.6x0.63 0.7x0.35
7 = 0.378 = 0.245 - 0.378 1
0.6x0.63 0.7x0.49 0.8x0.35
8 = 0.378 = 0.343 = 0.280 0.378 1
0.6x0.72 0.7x0.63 0.8x0.49
9 = 0.432 = 0.441 = 0.392 0.441 2
 Stage 2 (Continued)
y2 p2(y2) * F3(x2-c2(y2))
x2 1 2 3 F2(x2) y 2*
0.6x0.81 0.7x0.63 0.8x0.63
10 = 0.486 = 0.441 = 0.504 0.504 3

(Note: For stage 2, min x2 = 5 as at least one

unit of component 1 would have been installed
and at least one unit of components 3 and 4
have to be installed costing at least $400;
similarly max x2 = 9)
 Stage 1
y1 p1(y1) * F2(x1-c1(y1))
x1 1 2 3 F1(x1)
y1*
0.5x0.441 0.6x0.378 0.8x0.378
10 = 0.2205 = 0.2268 = 0.3024 0.3024 3

Thus the optimal solution is:

y1 = 3, y2 = 1, y3 = 1, y4 = 3

And z = 0.3024