Chapter 9

Linear Programming

9.1 Systems of Linear Inequalities

9.2 Linear Programming Involving Two Variables
9.3 The Simplex Method: Maximization
9.4 The Simplex Method: Minimization
9.5 The Simplex Method: Mixed Constraints

9.1
 9.1 Systems of Linear Inequalities

Linear inequalities ( 線性不等式 ):
For example, 3x – 2y < 6 and x + y  6 are linear
inequalities involving two variables


Solution of an inequality:
An ordered pair (a, b) is a solution of an inequality in x
and y if the inequality is true when a and b are substituted
for x and y, respectively


The graph of an inequality:
The collection of all solutions of the inequality

9.2

Ex 1:
Sketch the graph of the inequalities (a) x > – 2 and (b) y  3
Sol:
Step 1: Replace the inequality sign with an equal sign, and
sketch the graph of the corresponding equation, e.g., x = – 2
Step 2: Test one points in each of the regions separated by the
corresponding equation in step 1. If the point satisfies the
inequality, then shade the entire region to denote that every
point in the region satisfies the inequality

Notes: The corresponding equation separates the plane into
two regions, and one of the following two statements is true
(1) All points in the region are solutions of inequality
(2) No points in the region are solutions of the inequality 9.3
※ It is common to test the point of origin due to the least
computational effort

The solutions of x > –2 are The solutions of y  3 are

points lying to the right of x points lying below (or on)
= –2 (A dashed line is used the line of y = 3 (A solid
for > or <) line is used for  or )

9.4

Ex 2:
Sketch the graph of the inequalities x – y < 2

Sol:
※ Draw the graph of the
corresponding equation x – y =
2 first
※ Because the origin (0, 0)
satisfies the inequality, the
solution region is the half-
plane lying above (or to the
left of) the line x – y = 2

9.5

Systems of linear inequalities:
For example, x  y 12
3x  4 y 15
x 0
y 0

is a system of linear inequalities


Solution of a system of inequalities:
A solution is an ordered pair (x, y) satisfying each
inequality in the system

The graph of a system of inequalities:
The collection of all solutions of the system
9.6

Ex 3: Solving a system of inequalities
Sketch the graph and label the vertexes ( 頂點 ) of the solution
set of the system shown below
x y 2
x2
y 3

Sol:

Vertex A(–2, –4): the intersection of x – y = 2 and x = –2

Vertex B(5, 3): the intersection of x – y = 2 and y = 3
Vertex C(–2, 3): the intersection of x = –2 and y = 3

9.7

Notes: For complicated regions, two border lines could
intersect at a point that is not a vertex of the solution region

9.8

Notes: The system might have no solution or the solution
set of a system can be unbounded

No solution Solution region is

unbounded
x y 3 x  2y  3
x y 1 x y 3

9.9
Keywords in Section 9.1:

linear inequalities: 線性不等式

solution of an inequality: 線性不等式的解

systems of linear inequalities: 線性不等式系統

solution of a system of inequalities: 線性不等式系統的解

vertexes: 頂點

9.10
 9.2 Linear Programming Involving Two Variables

Linear programming problems ( 線性規劃問題 )
The linear programming problem is a type of optimization
problem consisting of a linear objective function ( 線性目標
函數 ) and a system of linear inequalities called
constraints

Note: The objective function gives the quantity that is to be
maximized (or minimized), and the constraints determine the
set of feasible solutions
※ Many applications in business and economics involve a process
called optimization ( 最佳化 ), e.g., to find the minimum cost, the
maximum profit, or the minimum consumption of resources
※ The linear programming problem is one of the simplest and most
widely used optimization problems
9.11

Ex 1: Solving a linear programming problem
Find the maximum value of z = 3x + 2y subject to the
constraints listed below x 0
y 0
x  2 y 4
x  y 1
Sol:
At (0, 0): z = 3(0) + 2(0) = 0
At (1, 0): z = 3(1) + 2(0) = 3
At (2, 1): z = 3(2) + 2(1) = 8 (maximum value of
z)
At (0, 2): z = 3(0) + 2(2) = 4

※ Try testing some interior points in the

region. You will see that corresponding
value of z are less than 8
9.12

Why the maximum value of the objective function must occur at
a vertex?
Consider to rewrite the objective function in the form
3 z
y  x 
2 2

※ First, the above equation represents a

family of parallel lines given different
values of z, each of which is with the
slope to be –3/2
※ Second, if you want to find the
maximum value of z for the solution
region, it is equivalent to find the line
with the largest y-intercept
※ Finally, it is obvious that such a line
must pass through one (or more) of the
vertexes of the solution region 9.13

Note: The linear programming problem involving two variables
can be solved by the graphical method ( 圖示解法 ):
1. Sketch the region corresponding to the system of constraints
(The points inside or on the boundary of the region are called feasible
solutions)
2. Find the vertexes of the region
3. Test the objective function at each of the vertexes and select
the values of the variables that optimize the objective function
(For a bounded region, both a minimum and maximum value will exist
at vertexes. For an unbounded region, if an optimal solution exists, it
must occur at a vertex)

9.14

Ex 2: Solving a linear programming problem
Find the maximum value of z = 4x + 6y, where x  0 and y 
0, subject to the constraints
 x  y 11
x  y 27
2 x  5 y 90
Sol:
At (0, 0): z = 4(0) + 6(0) = 0
At (0, 11): z = 4(1) + 6(11) = 66
At (5, 16): z = 4(5) + 6(16) = 116
At (15, 12): z = 4(15) + 6(12) = 132
At (27, 0): z = 4(27) + 6(0) = 108
So the maximum value of z is 132, and
this occurs when x = 15 and y = 12
9.15

Ex 3: Minimizing an objective function
Find the minimum value of the objective function
z = 5x + 7y,
where x  0 and y  0, subject to the constraints

2 x  3 6
3 x  y 15
 x  y 4
2 x  5 y 27

9.16
Sol:

At (0, 2): z = 5(0) + 7(2) = 14

At (0, 4): z = 5(0) + 7(4) = 28
At (1, 5): z = 5(1) + 7(5) = 40
At (6, 3): z = 5(6) + 7(3) = 51
At (5, 0): z = 5(5) + 7(0) = 25
At (3, 0): z = 5(3) + 7(0) = 15

So the minimum value of z is 14, and

this occurs when x = 0 and y = 2

9.17

Ex: The maximum (or minimum) value occurs at two different
vertexes
To maximize the objective function z = 2x + 2y given the
following shaded region

At (0, 0): z = 2(0) + 2(0) = 0

At (0, 4): z = 2(0) + 2(4) = 8
At (2, 4): z = 2(2) + 2(4) = 12 (maximum value of
z)
At (5, 1): z = 2(5) + 2(1) = 12 (maximum value of
z)
At (5, 0): z = 2(5) + 2(0) = 10

The maximum value occurs not only at the

vertexes (2, 4) and (5, 1), but also occurs at
any point on the line segment connecting
these two vertexes
9.18

Ex: An unbounded region
To maximize the objective function z = 4x + 2y given the
following shaded region

For this unbounded region, there is no

maximum value of z. By choosing
points (x, 0) for x  4, you can obtain
values of z = 4(x) + 2(0) = 4x that are
as large as you want

9.19

Ex 5: An application: minimum cost
The liquid portion of a diet is to provide at least 300 calories, 36
units of vitamin A, and 90 units of vitamin C daily. A cup of
dietary drink X provides 60 calories, 12 units of vitamin A, and
10 units of vitamin C. A cup of dietary drink Y provides 60
calories, 6 units of vitamin A, and 30 units of vitamin C.
Now suppose the dietary drink X cost $.12 per cup and drink Y
costs $.15 per cup. How many cups of each drink should be
consumed each day to minimize the cost and still meet the stated
daily requirements?

9.20
Sol: For calories: 60 x  60 y 300 
For vitamin A: 12 x  6 y 36 

For vitamin C: 10 x  30 y 90  constraints
x 0 

y 0 

The cost C : C 0.12 x  0.15 y objective function

At (0, 6): C = 0.12(0) + 0.15(6) = 0.90

At (1, 4): C = 0.12(1) + 0.15(4) = 0.72
At (3, 2): C = 0.12(3) + 0.15(2) = 0.66
At (9, 0): C = 0.12(9) + 0.15(0) = 1.08
So the minimum cost is $.66 per day, and this
occurs when 3 cups of drink X and 2 cups of
drink Y are consumed each day 9.21
Keywords in Section 9.2:

linear programming problems: 線性規劃問題

graphical method: 圖示解法

9.22
 9.3 The Simplex Method: Maximization

The graphical method is convenient to solve linear
programming problems involving only two variables. However,
for problems involving more than two variables or problems
involving large numbers of constraints, it is better to use a more
systematic solution, which is called the simplex method

Slack variables (for smaller-or-equal-to constraints) ( 鬆弛變數 )

 x1  x2 11  x1  x2  s1 11
x1  x2 27  x1  x2  s2 27
2 x1  5 x2 90 2 x1  5 x2  s3 90

The nonnegative numbers s1, s2, and s3 are called slack variables
because they represent the “slackness” in each inequality
9.23

Standard form of a maximization problem
A linear programming problem is in standard form if it
seeks to maximize the objective function z = c1x1 + c2x2 + · · ·
+ cnxn subject to the smaller-than-or-equal-to constraints
a11 x1  a12 x2    a1n xn  b1
a21 x1  a22 x2   a2 n xn  b2
 
am1 x1  am 2 x2   amn xn  bm

where bi  0 and xi  0. After adding slack variables, the

corresponding system of constraint equations is
a11 x1  a12 x2    a1n xn  s1  b1
a21 x1  a22 x2   a2 n xn  s2  b2
 
am1 x1  am 2 x2   amn xn  sm  bm
9.24

Note: For the standard form of a maximization problem, the
constraints involve only the smaller-than-or-equal-to signs
and all xi  0 and all bi  0


Basic solution ( 基本解 ):
A basic solution is a solution (x1, x2,…, xn, s1, s2,…, sm) for a
linear programming problem in standard form in which at
most m variables are nonzero


The variables that are nonzero are called basic variables ( 基本
變數 ), and the variables that are zero are called nonbasic
variables ( 非基本變數 )

9.25

Simplex tableau ( 單形法表格 ):
This tableau consists of the augmented matrix
corresponding to the constraint equations together with
slack variables and the objective function in the following
form  c1 x1  c2 x2    cn xn  (0)s1  (0)s2    (0)sm  z 0

Ex: z 4 x1  6 x2 objective function to be maximized
 x1  x2  s1  11 

x1  x2  s2  27  constraints
2 x1  5 x2  s3  90 
The corresponding simplex tableau is

current z-value 9.26


Note: For this initial simplex tableau, the basic variables
are s1, s2, and s3, and the nonbasic variables are x1 and x2, i.e.,
x1 = x2 = 0. As a consequence, the coefficient 1’s for s1, s2, and
s3 are similar to leading 1’s and these basic variables have
initial values of s1 = 11, s2 = 27, and s3 = 90

Note: The initial solution for the simplex tableau is
(x1, x2, s1, s2, s3) = (0, 0, 11, 27, 90)
and for the objective function, the current z-value is 4(0) +
6(0) = 0

Optimality check: Observing the bottom row of the tableau, if
any of these entries (except the current z-value) are negative,
the current solution is not optimal (so the initial values for (x1,
x2, s1, s2, s3) are not optimal) 9.27

Iteration of Simplex Method: An improvement process for finding
larger z-value, which brings a new nonbasic variable into the
solution, the entering variable ( 進入變數 ), and in the meanwhile
one of the current basic variables (the departing variable ( 離開變
數 )) must leave
1. The entering variable corresponds to the smallest (the most
negative) entry in the bottom row of the tableau
2. The departing variable corresponds to the smallest
nonnegative ratio of bi/aij in the column j determined by the
entering variable
3. The entry in the simplex tableau in the entering variable’s
column and the departing variable’s row is called the pivot entry
4. Use elementary row operations so that the pivot entry is 1, and
all other entries in the entering variable column are 0
9.28

Ex 1: Pivoting to find an improved solution
Use the simplex method to find an improved solution for the
linear programming problem represented by the tableau shown
below
※ Since –6 is the smallest
entry in the bottom row, we
choose x2 to be the entering
variable
※ Because the objective
function is z = 4x1 + 6x2, a
unit increase in x2 produce
an increase of 6 in z,
whereas a unit increase in x1
produces an increase of only
4 in z. Thus, adjusting x2
brings larger improvement
for finding a larger z-value
9.29
To find the departing variable, identify the row with the smallest
nonnegative bi/aij for the entering variables column, i.e., for j = 2
and examine b1/a12, b2/a22, and b3/a32
11 27 90
11, 27, and 18
1 1 5

(Since 11 is the
smallest nonnegative
ratio, we choose s1 as
the departing variable)

9.30
 G.-J. E. for
the 2nd column
    

This process is called pivoting

Therefore, the new tableau now becomes as follows


Note: x2 has replaced s1 to be a basic variable and the improved
solution is (x1, x2, s1, s2, s3) = (0, 11, 0, 16, 35) with a z-value of
9.31
※ The improved solution is not yet optimal because the bottom
row still has a negative entry
※ Since –10 is the smallest number in the bottom row, we choose
x1 as the entering variable
※ Moreover, the smallest nonnegative ratio among 11/(–1) = –11,
16/2 = 8, and 35/7 = 5 is 5, so s3 is the departing variable and
a31 is the pivot entry

pivot entry

9.32
So, the new tableau is as follows

※ (x1, x2, s1, s2, s3) = (5, 16,

0, 6, 0) with a z-value
of 4x1 + 6x2 = 4(5) +
6(16) = 116

Following the same rule, we can identify s1 to be the entering

variable and s2 to be the departing variable

9.33
By performing one more iteration of the simplex method, you
can obtain the following tableau

In this tableau, there are no negative elements in the bottom

row. So, the optimal solution is determined to be
(x1, x2, s1, s2, s3) = (15, 12, 14, 0, 0)
with
z = 4x1 + 6x2 = 4(15) + 6(12) = 132
9.34

Note: For the linear programming problem in Example 1
involving two variables, you could also employ the graphical
method to solve this problem as follows

※ Note that each iteration in the simplex method corresponds to moving from
a given vertex to an adjacent vertex with an improved z-value
(0, 0) (0,11) (5,16) (15,12)
  
z 0 z 66 z 116 z 132
※ In addition, why we choose (x1, x2, s1, s2, s3) = (0, 0, b1, b2, b3) = (0, 0, 11,
27, 90) as the initial basic solution is that (x1, x2) = (0, 0) must be a vertex
of the feasible solution region because all xi  0 in the standard form of
9.35

(x1, x2) = (0, 0) (intersection of x = 0 and y = 0)
  x1  x2  (0)  (0) 0 11  s1 11
 x 0 
nonbasic:  1 basic:  x1  x2 (0)  (0) 0 27  s2 27
 x2 0 2 x  5 x 2(0)  5(0) 0 90  s 90
 1 2 3


(x1, x2) = (0, 11) (intersection of x = 0 and –x + y = 11)
 x2 11
 x1 0 
nonbasic:  basic:  x1  x2 (0)  (11) 11 27  s2 16
 s1 0   x1  x2 11 2 x  5 x 2(0)  5(11) 55 90  s 35
 1 2 3


(x1, x2) = (5, 16) (intersection of –x + y = 11 and 2x + 5y = 90)
 x1 5
 s 0   x1  x2 11 
nonbasic:  1 basic:  x2 16
 s3 0  2 x1  5 x2 90  x  x (5)  (16) 21 27  s 6
 1 2 2


(x1, x2) = (15, 12) (intersection of x + y = 27 and 2x + 5y = 90)
 x1 15
 s 0  x1  x2 27 
nonbasic:  2 basic:  x2 12
 s3 0  2 x1  5 x2 90  x  x  (15)  (12)  3 11  s 14
 1 2 1

※ The above analysis shows that two nonbasic variables represent two straight lines (corresponding
equations) that determine the vertex we examine
※ The intuition of the simplex method is no more than testing different combinations of nonbasic
variables, which is equivalent to testing different vertexes of the solution region 9.36

Ex 2: The simplex method with three variables
Use the simplex method to find the maximum value of
z = 2x1 – x2 + 2x3,
subject to the constraints
2 x1  x2  10
x1  2 x2  2 x3  20
x2  2 x3  5

where x1  0, x2  0, and x3  0

Sol:
First, construct the initial simplex tableau with the initial
basic solution (x1, x2, x3, s1, s2, s3) = (0, 0, 0, 10, 20, 5)

9.37
※ Note that the “tie” occurs when choosing the
entering variable. Here we arbitrarily choose
x3 as the entering variable

9.38
 So the optimal solution is
(x1, x2, x3, s1, s2, s3) = (5, 0, 5/2, 0, 20, 0)
with
z = 2x1 – x2 + 2x3 = 2(5) – (0) + 2(5/2) = 15


Note: Since s1 = s3 = 0, the optimal solution satisfies the first
and third constraints. The result of s2 = 20 indicates that the
second constraint has a slack of 20 9.39

Ex 3: The simplex method with equality constraints
Use the simplex method to find the maximum value of
z = 3x1 + 2x2 + x3,
subject to the constraints
4 x1  x2  x3  30
2 x1  3 x2  x3  60
x1  2 x2  3 x3  40

where x1  0, x2  0, and x3  0
Sol: ※ The initial basic solution is (x1, x2, x3,
s1, s2, s3) = (0, 0, 0, 30, 60, 40)

※ The variable s1 is an artificial variable,

rather than a slack variable
※ Since the first constraint is an equality, s1
in the optimal solution must be zero, i.e.,
s1 must be the nonbasic variable and thus
be 0 in the optimal solution 9.40
9.41
 So the optimal solution is
(x1, x2, x3, s1, s2, s3) = (3, 18, 0, 0, 0, 1)
with
z = 3x1 + 2x2 + x3 = 3(3) + 2(18) + 1(0) = 45


Note: Since s1 is not a basic variable and thus s1 = 0 in the
optimal solution, this solution satisfies the first constraints (4(3)
+ 1(18) + 1(0) = 30) 9.42

Ex 4: A business application: maximum profit
A manufacturer produces three types of plastic fixtures ( 配件 ).
The times required for molding ( 鑄模 ), trimming ( 修整 ), and
packaging are presented as follows. (Times and profits are
displayed in hours and dollars for per dozen fixtures)

How many dozen of each type of fixture should be produced to

obtain a maximum profit?
9.43
Sol:
Let x1, x2, x3 represent the numbers of dozens of types A, B, and
C fixtures, respectively. The objective function is expressed as
profit = P = 11x1 + 16x2 + 15x3,
subject to the constraints
2
x1  2 x2  x3  12000
3
2 2
x1  x2  x3  4600
3 3
1 1 1
x1  x2  x3  2400
2 3 2

where x1  0, x2  0, and x3  0

9.44
Apply the simplex method with the initial basic solution (x1, x2,
x3, s1, s2, s3) = (0, 0, 0, 12000, 4600, 2400)

9.45
So the optimal solution is (x1, x2, x3, s1, s2, s3) = (600, 5100,
800, 0, 0, 0) with the maximum profit to be $100,200 9.46

Ex 5: A business application: media selection
The advertising alternatives for a company include television,
radio, and newspaper advertisements. The costs and estimates
of audience coverage are provided as follows

Moreover, in order to balance the advertising among the three

types of media, no more than half of the total number of
advertisements should occur on the radio, no more than 10
advertisements should occur on the newspaper, and at least
10% should occur on the television. The weekly advertising
budget is $18,200. How many advertisements should be run on
each media to maximize the total audience? 9.47
Sol:
Let x1, x2, x3 represent the numbers of advertisements in
television, newspaper, and radio, respectively. The objective
function is expressed as
z = 100000x1 + 40000x2 + 18000x3,
where x1  0, x2  0, and x3  0, and subject to the constraints
2000 x1  600 x2  300 x3  18200
x2  10
x3  0.5( x1  x2  x3 )
x1  0.1( x1  x2  x3 )

20 x1  6 x2  3x3  182
x2  10

 x1  x2  x3  0
 9 x1  x2  x3  0 9.48
Apply the simplex method with the initial basic solution (x1, x2,
x3, s1, s2, s3, s4)= (0, 0, 0, 182, 10, 0, 0)

9.49
So the optimal solution is (x1, x2, x3, s1, s2, s3, s4) = (4, 10, 14, 0,
0, 0, 12) with the maximum weekly audience to be 1,052,000 9.50
Keywords in Section 9.3:

simplex method: 單形法

slack variable: 鬆弛變數

basic solution: 基本解

basic variable: 基本變數

nonbasic variable: 非基本變數

simplex tableau: 單形法表格

entering variable: 進入變數

departing variable: 離開變數

artificial variable: 人工變數

9.51
 9.4 The Simplex Method: Minimization

In Section 9.3, the simplex method is applied only to linear
programming problems in standard form where the objective
function is maximized. In this section, we will study the
minimization problem

Standard form for a minimization problem
If w = c1x1 + c2x2 + · · · + cnxn is to be minimized, subject to
the constraints
a11 x1  a12 x2    a1n xn  b1
a21 x1  a22 x2   a2 n xn  b2
 
am1 x1  am 2 x2   amn xn  bm

where xi  0 and bi  0 Note that in the standard form for

maximization problems, those are
“” signs (see Slide 9.24) 9.52

To find the Dual Maximization Problem (corresponding to the
minimization problem that we want to solve)
Considering Example 5 in Section 9.2, we want to minimize
w 0.12 x1  0.15 x2 , subject to the constraints
60 x1  60 x2 300
12 x1  6 x2 36
10 x1  30 x2 90
where x1  0 and x2  0

※ We cannot apply directly the simplex method because the simplex method
introduced in Section 9.3 is for maximization problems
※ There is a dual relationship between minimization problems in standard
form and maximization problems in standard form

9.53
The steps to convert this problem into a maximization problem
1. Form the augmented matrix as follows


 constraints for the minimization problem



 objective function for the minimization problem

2. Find the transpose of this matrix


 constraints for the dual maximization problem


 objective function for the dual maximization problem

3. Interpret the transposed matrix as a maximization problem

9.54
※ The corresponding maximization problem is called the dual
maximization problem of the original minimization problem,
which is as follows
Find the maximum value of z 300 y1  36 y2  90 y3 ,
subject to the constraints
60 y1  12 y2  10 y3 0.12
60 y1  6 y2  30 y3 0.15
where y1 0, y2 0, and y3 0

※ The solution of the original minimization problem can be found

by applying the simplex method to the new dual maximization
problem

9.55
Apply the simplex method with the initial basic solution (y1, y2,
y3, s1, s2)= (0, 0, 0, 0.12, 0.15)

9.56
So the optimal solution is (y1, y2, y3, s1, s2) = (7/4000, 0, 3/2000, 0,
0) with the maximum value to be 300(7/4000)+36(0)+90(3/2000)
= 33/50 = 0.66
※ The optimal values of xi’s for the original minimization can be obtained
from the entries in the bottom row corresponding to slack variable columns,
i.e., x1 = 3 and x2 = 2
※ The minimum value for the original objective function is 0.12(3)+0.15(2) =
0.66, which is exactly the maximum value for the dual maximization
problem (It is the same as that derived by the graphical method in Sec. 9.2)
※ This dual relationship is called the von Neumann Duality Principle,
introduced by the American mathematician John von Neumann (1903- 9.57

Ex 1: Solving a minimization problem
Find the minimum value of w = 3x1 + 2x2, subject to the
constraints
2 x1  x2  6
x1  x2  4

where x1  0 and x2  0
Sol:
1. The augmented matrix corresponding the this
minimization problem

9.58
2. Derive the matrix corresponding to the dual maximization
problem by taking transpose

The above augmented matrix implies the following

maximization problem:
Find the maximum value of z = 6y1+ 4y2, subject to the
constraints
2 y1  y2  3
y1  y2  2
where y1  0 and y2  0

9.59
Apply the simplex method with the initial basic solution (y1, y2,
s1, s2)= (0, 0, 3, 2)

9.60
From this final simplex tableau, you can see the
maximum value of z is 10. So the solution of the
original minimization problem is w =10, and this
occurs when (x1, x2) = (2, 2)

9.61

Ex 2: Solving a minimization problem
Find the minimum value of w = 2x1 + 10x2 + 8x3, subject to the
constraints
x1  x2  x3  6
x2  2 x3  8
 x1  2 x2  2 x3  4

where x1  0, x2  0, and x3  0
Sol:
First, transpose the augmented matrix

 transpose
  

9.62
The corresponding maximization problem:
Find the maximum value of z = 6y1 + 8y2 + 4y3, subject
to the constraints
y1  y3  2
y1  y2  2 y3  10
y1  2 y2  2 y3  8

where y1  0, y2  0, and y3  0

The initial simplex tableau

9.63
From the last row of this final simplex tableau, you can see the
maximum value of z is 36. So the solution of the original minimization
problem is w =36, and this occurs when (x1, x2 , x2) = (2, 0, 4) 9.64

Ex 3: A business application: minimum cost
A small petroleum company owns two refineries. Refinery 1
costs $20,000 per day to operate, and it can produce 400 barrels
of high-grade oil, 300 barrels of medium-grade oil, and 200
barrels of low-grade oil each day. Refinery 2 is newer and more
modern. It costs $25,000 per day to operate, and it can produce
$300 barrels of high-grade oil, 400 barrels of medium-grade
oil, and 500 barrels of low-grade oil each day
The company has orders totaling 25,000 barrels of high-
grade oil, 27,000 barrels of medium-grade oil, and 30,000
barrels of low-grade oil. How many days should this petroleum
company run each refinery to minimize its costs and still refine
enough oil to meet its orders?

9.65
Sol:
Let x1 and x2 represent the numbers of days on which the two
refineries are operated. Then the total cost is represented by
C = 20,000x1 + 25,000x2,
and the constraints are
400 x1  300 x2  25, 000 (high-grade oil)
300 x1  400 x2  27, 000 (medium-grade oil)
200 x1  500 x2  30, 000 (low-grade oil)

where x1  0 and x2  0

 transpose
  

9.66
Now apply the simplex method to the dual problem

9.67
From this final simplex tableau, you can see the
solution for the original minimization problem is C =
$ 1,750,000 and this occurs when x1 = 25 and x2 =
50, i.e., the two refineries should be operated for 25
and 50 days, respectively

9.68
Keywords in Section 9.4:

dual maximization problem: 對偶極大化問題

9.69
 9.5 The Simplex Method: Mixed Constraints

In Sections 9.3 and 9.4, we learn how to solve linear
programming problems in the standard forms, i.e., the
constraints for the maximization problem involve only “”
inequalities, and the constraints for the minimization problem
involve only “” inequalities

Mixed-constraint problems: the constraints involve both
types of inequalities

For example, find the maximum value of z = x1 + x2 + 2x3,
subject to the constraints
2 x1  x2  x3  50 Because of these two
 inequalities, this is
2 x1  x2  36
not the standard form
x1  x3  10 for a maximization
problem
where x1  0, x2  0, and x3  0 9.70

For the first inequality, it does involve  inequalities, so we
add a slack variable to form the equation as before
2 x1  x2  x3  s1  50

For the last two inequalities, a new type of variable, called a
surplus variable ( 剩餘變數 ), is introduced as follows
2 x1  x2  s2  36
x1  x3  s3 = 10


Note: The variable s2 and s3 are called the surplus variables
because they represent the amounts by which the left sides of
the inequalities exceed the right sides

Note: The surplus variables as well as the slack variables must
be nonnegative, i.e., s1  0, s2  0, and s3  0 in the above case
9.71

The initial simplex tableau can be formed as follows


Note: According to the rules introduced in the Section 9.3, the
initial basic solution is (x1, x2, x3, s1, s2, s3) = (0, 0, 0, 50, –36, –
10), which is not a feasible solution because the values of the two
surplus variables are negative

In other words, the point (x1, x2, x3) = (0, 0, 0) is not an vertex of
the feasible solution region (In fact, (0, 0, 0) is not even in the
feasible solution region)

So, we cannot apply the simplex method directly from this initial
basic solution 9.72

We need to use the “trial and error” method ( 試誤法 ) to find a
feasible solution for the initial basic solution for the simplex
method, that is, arbitrarily choosing new entering variable to
replace the negative surplus variables (It is equivalent to jump from the
intersection of x1 = 0, x2 = 0, and x3 = 0 to the intersection of other three
corresponding equations)


Note: Here we choose x3 as the entering variable and s3 as the
departing variable arbitrarily, i.e., we do not follow the pivoting
process introduced on Slide 9.28 to select the entering and
departing variables 9.73

After pivoting, the new simplex tableau is as follows

※ It means that we consider the intersection of x1 = 0, x2 = 0, and x1 + x3 = 10

(because s3 = 0)

The current solution (x1, x2, x3, s1, s2, s3) = (0, 0, 10, 40, –36, 0) is
not feasible because the value of s2 is still negative

So we choose x2 as the entering variable and s2 as the departing
variable and pivot to obtain the following simplex tableau 9.74
 ※ The above tableau suggests that we consider the intersection of x1 = 0, 2x1 +
x2 = 36 (because s2 = 0), and x1 + x3 = 10 (because s3 = 0)

We finally obtain a feasible solution, which is (x1, x2, x3, s1, s2, s3)
= (0, 36, 10, 4, 0, 0)

From here on, you can apply the simplex method as usual, i.e.,
we choose s3 as the entering variable because of its smallest entry
in the bottom row, and then compare bi/aij in the entering variable
column to find s to be the departing variable 9.75

After pivoting, you can obtain the following simplex tableau


Note that it is the final tableau because it represents a feasible
solution and there are no negative entries in the bottom row

So, you can conclude that the maximum value of z = 64 and this
occur when (x1, x2, x3, s1, s2, s3)= (0, 36, 14, 0, 0, 4)

9.76

For the minimization problem with mixed-constraints, the
situation is more difficult because we cannot apply the dual
principle to find the corresponding maximization problem if there
are mixed constraints

Another technique is introduced to change a mixed-constraint
minimization problem to a mixed-constraint maximization
problem by multiplying each coefficient in the objective function
by –1

This technique is demonstrated in the next example

9.77

Ex 2: A minimization problem with mixed constraints
Find the minimum value of w = 4x1 + 2x2 + x3, subject to the
constraints
2 x1  3 x2  4 x3  14
3 x1  x2  5 x3  4
x1  4 x2  3 x3  6

where x1  0, x2  0, and x3  0
Sol:
First, rewrite the objective function by multiplying each of
its coefficients by –1, i.e., z = –4x1 – 2x2 – x3 (maximizing
this new objective function is equivalent to minimizing
the original objective function)

9.78
Next, add a slack variable to the first inequality and subtract
surplus variables from the second and third inequalities, and the
initial simplex tableau is as follows

※ Note that the bottom row contains the negatives of the coefficients of the
new objective function
※ The current solution (x1, x2, x3, s1, s2, s3) = (0, 0, 0, 14, –4, –6) is not a
feasible solution
※ So, what we need to do is to arbitrarily choose the entering and departing
variables to find a feasible solution
※ Here we choose x2 as the entering variable and s2 as the departing variable 9.79
※ The current solution (x1, x2, x3, s1, s2, s3) = (0, 4, 0, 2, 0, 10) is a feasible
solution. From now on, we can apply the standard simplex method
※ We choose x3 as the entering variable because –9 is the smallest entry in the
bottom row and choose s3 as the departing variable because b3/a33 is the
smallest nonnegative ratio in the third column
After pivoting, we can obtain the following simplex tableau

9.80
After pivoting around a25, we can obtain the following simplex
tableau

The maximum value of the new objective function is z = –2, so

the minimum value of the original objective function is w = 2.
This occurs when x1 = 0, x2 = 0, and x3 = 2

※ The solution of x3 is according to the third row since it is a basic

variables, and x1 = 0 and x2 = 0 are due to the fact that they are nonbasic
variables
9.81

Ex 3: A business application: minimum shipment costs
An automobile company has two factories. One factory has 400
cars (of a certain model) in stock and the other factory has 300
cars (of the same model) in stock. Two customers order this car
model. The first customer needs 200 cars, and the second
customer needs 300 cars. The costs of shipping cars from the
two factories to the customers are shown in the following table.

How should the company ship the cars in order to minimize the
shipping costs?

9.82
 Sol:
Let x1 and x2 represent the numbers of cars shipped from factory
1 to the first and second customers, respectively. The total cost
of shipping to be minimized is
C = 30x1 + 25x2 + 36(200 – x1) + 30(300 – x2)
= 16,200 – 6x1 – 5x2
and the constraints are
x1  x2 400 (There are only 400 cars in factory 1)
(200  x1 )  (300  x2 ) 300 (There are only 300 cars in factory 2)
x1 200 (The max. number of cars from factory 1 to customer 1 is 200)
x2 300 (The max. number of cars from factory 1 to customer 2 is 300)

where x1  0 and x2  0 and the second inequality can be

rewritten as x1 + x2  200
9.83
Multiply the objective function by –1 to convert it into a
maximization problem, i.e., to maximize z = 6x1 + 5x2 – 16,200.
So, the initial simplex tableau is as follows

※ The current solution (x1, x2, s1, s2, s3, s4) = (0, 0, 400, –200, 200, 300) and
the current z-value is –16,200
※ Since the current solution is not a feasible solution (due to the negative
value of s2), in addition to choosing s2 as the departing variable, we
arbitrarily choose x1 to be the entering variable 9.84
After pivoting, we can obtain the following simplex tableau

※ The current solution (x1, x2, s1, s2, s3, s4) = (200, 0, 200, 0, 0, 300) is a
feasible solution
※ So, from now on, we can apply the standard simplex method. Following
the rule on Slide 9.28, we choose s2 as the entering variable and s3 as the
departing variable
9.85
After pivoting, we can obtain the following simplex tableau

※ Following the rule on Slide 9.28, we choose x2 as the entering variable

and s1 as the departing variable. After pivoting around the entry a12, we
can obtain the following simplex tableau

9.86
※ Since all entries except the z-value in the bottom row are nonnegative, it
is the final simplex tableau
※ The minimum shipping cost is $14,000, and this occurs when x1 = 200
and x2 = 200, i.e., the numbers of cars shipped from the two factories
are as follows

9.87
 Linear Programming under xi  0 and bi 
0 problems in standard
Max. Min. problems in standard
form, i.e., involving only dual transformation
form, i.e., involving only
“” in constraints “” in constraints
1. Add slack variables to There are two kinds of methods to
form constraint equations solve this problem
2. Perform the simplex
method

Max. problems with mixed Min. problems with mixed

constraints constraints
1. Add slack variables and 1. Multiply the objective function
subtract surplus variables by –1 to convert it into a max.
to form constraint problem
equations 2. Add slack variables and
2. Try to find a feasible subtract surplus variables to
solution by trial and error form constraint equations
3. Apply the simplex method 3. Try to find a feasible solution
by trial and error
4. Apply the simplex method 9.88

Ex in Section 9.4 on Slide 9.53: Do not use the dual principle to
solve the minimization problem in standard form
We want to minimize w 0.12 x1  0.15 x2 , subject to the
constraints
60 x1  60 x2 300
12 x1  6 x2 36
10 x1  30 x2 90

where x1  0 and x2  0
Sol:
First, rewrite the objective function by multiplying each of
its coefficients by –1, i.e., z = –0.12x1 – 0.15x2
(maximizing this new objective function is equivalent to
minimizing the original objective function)

9.89
Next, subtract surplus variables for these three inequalities, and
the initial simplex tableau is as follows
x1 x2 s1 s2 s3 b Basic Variables
60 60 1 0 0 300 s1
12 6 0 1 0 36 s2
10 30 0 0 1 90 s3
0.12 0.15 0 0 0 0

※ The current solution (x1, x2, s1, s2, s3) = (0, 0, –300, –36, –90) is not
a feasible solution
※ So, what we need to do is to arbitrarily choose the entering
variables and choose negative surplus variables as the departing
variables to find a feasible solution
※ Here we choose x2 as the entering variable and s2 as the departing
variable

9.90
 After pivoting, we can obtain the following simplex tableau
x1 x2 s1 s2 s3 b Basic Variables
 60 0 1 10 0  60 s1
1
2 1 0  0 6 x2
6
 50 0 0 5  1  90 s3
0.15
 0.18 0 0 0  0.9
6
※ The current solution (x1, x2, s1, s2, s3) = (0, 6, 60, 0, 90) is a feasible solution.
From now on, we can apply the standard simplex method
※ We choose x1 as the entering variable because –0.18 is the smallest (or the most
negative) entry in the bottom row except the z-value and choose s1 as the
departing variable because b1/a11 is the smallest nonnegative ratio in the first
column
※ After performing the pivoting process based on a11, we can obtain the following9.91
x1 x2 s1 s2 s3 b Basic Variables
1 1
1 0  0 1 x1
60 6
1 1
0 1  0 4 x2
30 6
5 20
0 0   1  40 s3
6 6
0.18 0.03
0 0  0  0.72
60 6
x1 x2 s1 s2 s3 b Basic Variables ※ Since all entries except the z-value
on the bottom row is nonnegative,
3 1
1 0 0 3 x1 it is the final simplex tableau
120 20 ※ The minimum value for w is 0.66,
1 1 and this occurs when x1 = 3 and x2
0 1 0  2 x2
120 20 =2
1 6 ※ This method can generate the same
0 0  1 12 s2 result as that in the method
4 20 combining the dual transformation
0.21 0.03 and the standard simplex method
0 0 0  0.66
120 20 (comparing to Slide 9.57) 9.92
Keywords in Section 9.5:

mixed-constraint problems: 混合限制問題

surplus variable: 剩餘變數

9.93

Solving the linear programming problem by the function
“Solver” in Excel (see Homework2_Portfolio_frontier.xls”)

Not only linear programming problems ( 線性規劃問題 ) but also
nonlinear programming problems can be solved by “Solver”

The nonlinear programming problem ( 非線性規劃問題 ) means the
objective function and/or one or more constraints are nonlinear

The quadratic programming problem ( 二次規劃問題 ) is a special
case of the nonlinear programming problem, in which the
objective function contains squared terms and the constraints are
all linear

Finding the portfolio frontier is a quadratic programming
problem

9.94

Portfolio frontier problem: Given an expected portfolio
return, find the optimal weights on assets to construct a
portfolio with minimum variance
Suppose the expected returns E(ri) and variances and
covariance among individual assets cov(ri, rj) are known.
Given any specified value of E(rp), the portfolio frontier
problem is to find optimal weights w1, w2,…, wn in the
following quadratic programming problem:
n n n n
min  P2   wi w j cov(ri , rj )   wi w j i j i , j
w1 ,, wn
i 1 j 1 i 1 j 1

s.t. w1 E (r1 )  w2 E (r2 )    wn E (rn ) E (rp )

w1  w2    wn 1
9.95

Homework 2
– Find the portfolio frontier based on at least 5 stocks in S&P 500,

i.e., solve the above quadratic programming problem for n  5

– Note that ri’s of all stocks are total returns, so you need to find adjusted
closing prices or adjusted returns
– Today is assumed to be Sept. 1, 2015 and you need to find one-year
historical returns for each stock
– Estimate the expected values and standard deviations for the rate of
returns of all stocks and the correlations among the rate of returns of
stocks as well, i.e., to calculate sample averages (for estimating E(ri)) ,
σi, and ρij from the historical data
– Annualization: daily average return × 252 = annual average return;
daily s.d. × sqrt(252) = annual s.d.
– For different portfolio returns, e.g., E(rp) = 0, 1%, 2%,…, 40%, solve wi
for each E(rp) by the function “Solver” in Excel
– For each solution set of wi, plot a point according to the portfolio return9.96
– Bonus: conduct further analyses, for example,
– Compare the portfolio frontiers of at least three industries (1 point)
– Compare the portfolio frontiers given different values of n (With the
increase of n, the portfolio frontier should move further toward left
theoretically, i.e., given the same E(rp), you can generate a smaller
portfolio standard deviation given more stocks) (1 point)
– Examine the special case for n = 2: (1 point)
– Check that the point of (E(r ), σ ) for each of the two stocks lies on
i i

the portfolio frontier

– Consider the case of ρ
12 = 1 (simply setting ρ12 to be 1) and

demonstrate that the portfolio frontier is a straight line in this case

– Consider the case of ρ
12 = –1 (simply setting ρ12 to be –1) and

demonstrate that the portfolio frontier is a saw-like, piecewise linear

curve and it is possible to construct a portfolio with zero standard
deviation
– The basic requirement is 7 points, and the bonus is worth at
9.97