You are on page 1of 5

Mathematical Programming

Solutions to Examples 4
1. Phase I problem is
minimize z
I
= R
1
+R
3
s.t. x
1
+x
2
+R
1
= 150
x
1
+s
2
= 40
x
2
s
3
+R
3
= 20
x
1
; x
2
; s
2
; s
3
; R
1
; R
3
0
Tableaux are
min x
1
x
2
s
3
R
1
1 1 0 150
s
2
1 0 0 40
R
3
0

1 -1 20
z
I
1 2 -1 170
min x
1
R
3
s
3
R
1
1 -1

1 130
s
2
1 0 0 40
x
2
0 1 -1 20
z
I
1 -2 1 130
min x
1
R
3
R
1
s
3
1 130
s
2
1 40
x
2
1 150
z
I
0 0
z -10 1500
Final tableau shows new bottom row for Phase II which is optimal. So (x

1
; x

2
) = (0; 150) :
2. Primal is min x
1
+x
2
s.t. x
1
= 1
x
2
= 1
x
1
; x
2
0
Dual is max y
1
+y
2
s.t. y
1
1
y
2
1
y
1
; y
2
u.r.s.
The unique basis for the primal is optimal, so dual simplex iterations stop immediately
min x
1
x
2
x
1
1 0 1
x
2
0 1 1
0 0 2
Convert the dual constraints to standard form. 2 iterations are needed for optimality.
max y
1
y
2
s
1
1 0 1
s
2
0 1 1
-1 -1
max s
1
y
2
y
1
1 0 1
s
2
0 1 1
1 -1 1
max s
1
s
2
y
1
1 0 1
y
2
0 1 1
1 1 2
1
3. Solution of primal via dual simplex iterations:
Min x
1
x
2
x
3
x
4
s
1
1 6 3 1 2
s
2

2 5 1 3 3
2 15 5 6 0
Min s
2
x
2
x
3
x
4
s
1

1
2

17
2

5
2

5
2

1
2
x
1
1
2

5
2
1
2

3
2
3
2
1 20 4 9 3
Min s
2
x
2
s
1
x
4
x
3
1
5
x
1
7
5

1
5

32
5

8
5
5
19
5
x

7
5
0
1
5
0

; z
0
=
19
5
Dual problem is
maximize 2y
1
+ 3y
2
s.t. y
1
+ 2y
2
2
6y
1
5y
2
15
3y
1
+ y
2
5
y
1
3y
2
6
y
1
; y
2
0
The optimal solution is (y

1
; y

2
) =

8
5
;
1
5

with value w

=
19
5
as predicted by duality theorem
(values appear in nal tableau under s
1
; s
2
columns).
Complementary slackness principle )s

1
= s

2
= 0 at primal optimum. Also w

2
; w

4
> 0 (2nd,
4th dual slacks) ) x

2
= x

4
= 0:
We solve the system
x
1
+ 3x
3
= 2
2x
1
+x
3
= 3
to conrm x

7
5
; 0;
1
5
; 0

.
4. From a graph we can see that constraints (2) and (4) for the dual in Q2 are geometrically
redundant.
Suppose we minimize s
i
s.t. the remaining constraints apply. If s

i
> 0; this shows that
there are no active points (satisfying the i
th
constraint exactly) that are within the feasible
region F
i
of the remaining constraints. Thus the i
th
constraint does not intersect F
i
, so is
geometrically redundant.
Consider min w
2
= 15 6y
1
+ 5y
2
subject to the constraints (1),(3),(4) of the dual to the
2
previous question.
y
1
y
2
w
1
1 2 2
w
3

3 1 5
w
4
1 -3 6
w
2
6 -5 15
w
3
y
2
w
1
1
3
y
1
5
3
w
4
13
3
w
2
-2 -7 5
showing that w

2
= 15 10 = 5 > 0 at the point y

5
3
; 0

:
5. P: min c
T
x
s.t. Ax = b
x 0
D: max y
T
b
s.t. y
T
A c
T
y u.r.s.
We suppose that A = (BjN) is (mn) so x is (n 1) ; y is (m1) : Let x
0
= (y
0
; 0) be an
optimal solution where y
0
= B
1
b > 0
The dual constraints are equivalently y
T
A + v
T
= c
T
where v 0 are dual slacks. (non-
degenerate). At optimum, the C-S conditions apply. Hence x
j
v
j
= 0 (j = 1; :::; n). x
B
> 0 )
rst m dual slacks are zero so y
T
0
B = c
T
B
. Hence y
T
= c
T
B
B
1
is the unique solution.
Non-degenerate P (or D) )D (or P) has unique solution. Conversely if P (or D) has multiple
solutions then D (or P) is degenerate.
6. Denote farmers possible treatments by T1: regenerate, T2: replant. Suppose the farmer
treats x
1
acres with T1 and x
2
acres with T2. The prots per acre are $40 for T1 and $70
for T2. Hence the farmers problem is
maximize 40x
1
+ 70x
2
s.t. x
1
+x
2
100
10x
1
+ 50x
2
4000
x
1
; x
2
0
Solution by simplex algorithm:
max x
1
x
2
I
s
1
1 1 100
s
2
10

50 4000
z -40 -70 0
max x
1
s
2
II
s
1
4
5

1
50
20
x
2
1
5
1
50
80
z 26
7
5
5600
max s
1
s
2
III
x
1
5
4

1
40
25
x
2

1
4
1
40
75
z 32
1
2
3
4
6250
Farmer should apply T1 over 25 acres, T2 over 75 acres. Maximum prot z
0
= $6; 250:
Note: compare the cost of applying a single Treatment option
T1 only: Cost = $1,000, Return = $4,000, Prot = $3,000
T2 only: Cost = $4,000 (80 acres), Return = $9,600, Prot = $5,600 (Tableau II)
3
The value of a unit increase in b
2
is the "shadow cost" or dual variable y
2
: We have the dual
values
y
T
0
= c
T
B
B
1
= (32:5; 0:75)
and so additional borrowing that costs not more than $0.75 interest is worthwhile.
A more detailed analysis is as follows.
z
0
= c
T
B

b = c
T
B
B
1
b
= y
T
0
b
Hence if y
T
0
= (y
1
; y
2
) and b = (0; )
T
we have
z
0
= y
2

From the nal tableau y


T
0
= (32:5; 0:75) so z
0
= 0:75: Increase in optimal prot is $0.75 per
$ borrowed. To calculate valid range of , need to examine tableau r.h.s. for feasibility.under
change b to b:
b = B
1
b =

1
40

1
40

!
so increase in prot valid as long as 25
1
40
0 i.e. $1000:
7. (a) Formulation as LP is
max 30x
1
+ 10x
2
+ 50x
3
s.t. 6x
1
+ 3x
2
+ 5x
3
45 Labour (hrs.)
3x
1
+ 4x
2
+ 5x
3
30 Materials
x
1
; x
2
; x
3
0
(b)
max x
1
x
2
x
3
I
s
1
6 3 5 45
s
2
3 4

5 30
30 10 50 0
max x
1
x
2
s
2
II
s
1

3 1 1 15
x
3
3
5
4
5
1
5
6
0 30 10 300
Tableau II is optimal, however "0" in bottom row indicates possible alternative solutions.
Pivot on x
1
column:
max s
1
x
2
s
2
III
x
1
1
3

1
3

1
3
5
x
3

1
5
1
2
5
3
0 30 10 300
There are two optimal (BFS) solutions (A) x

1
= (0; 0; 6) and (B) x

2
= (5; 0; 3) : Any convex
4
linear combination x

2
+ (1 ) x

1
= (5; 0; 6 3) is also optimal for 0 < < 1.
(c) Dual is
min 45y
1
+ 30y
2
s.t. 6y
1
+ 3y
2
30
3y
1
+ 4y
2
10
5y
1
+ 5y
2
50
y
1
; y
2
0
(d) Since the primal optimal solution is non-degenerate, the dual optimal solution must be
unique. We can conrm this by using C-S conditions. Applied to x

1
gives y

1
= 0 and
5y
1
+ 5y
2
= 50 ) y

2
= 10: The same result is obtained if C-S conditions are applied to x

2
:
NB Non-unique primal optimum =)degenerate dual optimum as predicted by Q5:
(e) Let c
1
= 30 + : Basis (A) remains optimal while
50
3
5
(30 + ) 0 ) 0
i.e. c
1
30: Basis (B) remains optimal while
1
3
(30 + ) 10 0

1
3
(30 + ) + 50 10 0

1
3
(30 + ) +
2
5
50 0
) 0 30 i.e. 30 c
1
60: The value of the optimal solution with Basis (B) varies
with since z
0
= 300 + 5:
For c
1
= 72; optimal solution becomes x

15
2
; 0; 0

with value 540


max s
1
x
2
s
2
IV
x
1
1
3

1
3

1
3
5
x
3

1
5
1
2
5
3
14 16 4 510
max s
1
x
2
x
3
V
x
1
1
3

1
3

1
3
15
2
s
2

1
5
1
2
5
15
2
12 26 10 540
(f) Value of additional kilo of materials is corresponding shadow price y

2
= 10: Consider
b = (0; )
T

b = B
1
b
=
"

1
5

#
for (A)
=
"

1
3

2
5

#
for (B)
valid while 15 0 from (A) and 5
1
3
0 from (B). Thus 15:
5

You might also like