4 Duality Theory
4 Duality Theory
=
=
n
j
j j
x c Z
1
,
=
=
m
i
i i
y b W
1
,
=
s
n
j
i j ij
b x a
1
,
=
>
m
i
j i ij
c y a
1
,
for for . , , 2 , 1 m i = . , , 2 , 1 n j =
for
. , , 2 , 1 m i =
for
. , , 2 , 1 n j = , 0 >
j
x , 0 >
i
y
In matrix notation
Primal Problem Dual Problem
Maximize
subject to
. 0 > x
. 0 > y
Minimize
subject to
b Axs
c yA>
, cx Z = , yb W =
Where and are row
vectors but and are column vectors.
c
| |
m
y y y y , , ,
2 1
=
b
x
Example
Max
s.t.
Min
s.t.
Primal Problem
in Algebraic Form
Dual Problem
in Algebraic Form
, 5 3
2 1
x x Z + =
, 18 12 4
3 2 1
y y y W + + =
18 2 3
2 1
s + x x
12 2
2
s x
4 s
1
x
0 x , 0 x
2 1
> >
5 2 2
3 2
> + y y
3 3
3
> + y
1
y
0 y , 0 y , 0 y
3 2 1
> > >
Max
s.t.
Primal Problem
in Matrix Form
Dual Problem
in Matrix Form
Min
s.t.
| | , 5 , 3
2
1
(
=
x
x
Z
(
(
(
s
(
(
(
(
18
12
4
,
2
2
0
3
0
1
2
1
x
x
.
0
0
2
1
(
>
(
x
x
| | | |. 0 , 0 , 0 , ,
3 2 1
> y y y
| | | | 5 , 3
2
2
0
3
0
1
, ,
3 2 1
>
(
(
(
y y y
| |
(
(
(
=
18
12
4
, ,
3 2 1
y y y W
Primal-dual table for linear programming
Primal Problem
Coefficient of: Right
Side
R
i
g
h
t
S
i
d
e
D
u
a
l
P
r
o
b
l
e
m
C
o
e
f
f
i
c
i
e
n
t
o
f
:
m
y
y
y
2
1
21
11
a
a
22
12
a
a
n
n
a
a
2
1
1
x
2
x
n
x
1
c
2
c
n
c
VI VI VI
Coefficients for
Objective Function
(Maximize)
1
b s
mn
a
2 m
a
1 m
a
2
b s
m
b s
C
o
e
f
f
i
c
i
e
n
t
s
f
o
r
O
b
j
e
c
t
i
v
e
F
u
n
c
t
i
o
n
(
M
i
n
i
m
i
z
e
)
One Problem Other Problem
Constraint Variable
Objective function Right sides
i
i
Relationships between Primal and Dual Problems
Minimization Maximization
Variables
Variables
Constraints
Constraints
0 >
0 >
0 s
0 s
s
s
>
>
=
=
Unrestricted
Unrestricted
The feasible solutions for a dual problem are
those that satisfy the condition of optimality for
its primal problem.
A maximum value of Z in a primal problem
equals the minimum value of W in the dual
problem.
Rationale: Primal to Dual Reformulation
Max cx
s.t. Ax b
x 0
L(X,Y) = cx - y(Ax - b)
= yb + (c - yA) x
>
s
Min yb
s.t. yA c
y 0
>
>
Lagrangian Function )] , ( [ Y X L
X
Y X L
c
c )] , ( [
= c-yA
The following relation is always maintained
yAx yb (from Primal: Ax b)
yAx cx (from Dual : yA c)
From (1) and (2), we have (Weak Duality)
cx yAx yb
At optimality
cx* = y*Ax* = y*b
is always maintained (Strong Duality).
>
s
s s
s
>
(1)
(2)
(3)
(4)
Complementary slackness Conditions are
obtained from (4)
( c - y*A ) x* = 0
y*( b - Ax* ) = 0
x
j
* > 0 y*a
j
= c
j
, y*a
j
> c
j
x
j
* = 0
y
i
* > 0 a
i
x* = b
i
, a
i
x* < b
i
y
i
* = 0
(5)
(6)
Any pair of primal and dual problems can be
converted to each other.
The dual of a dual problem always is the primal
problem.
Min W = yb,
s.t. yA c
y 0.
Dual Problem
Max (-W) = -yb,
s.t. -yA -c
y 0.
Converted to
Standard Form
Min (-Z) = -cx,
s.t. -Ax -b
x 0.
Its Dual Problem
Max Z = cx,
s.t. Ax b
x 0.
Converted to
Standard Form
s
s
>
>
> >
>
>
Min
s.t.
6 4 . 0 6 . 0
6 5 . 0 5 . 0
7 . 2 1 . 0 3 . 0
2 1
2 1
2 1
> +
= +
s +
x x
x x
x x
0 , 0
2 1
> > x x
2 1
5 . 0 4 . 0 x x +
Min
s.t.
] [y 6 4 . 0 6 . 0
] [y 6 5 . 0 5 . 0
] [y 6 5 . 0 5 . 0
] [y 7 . 2 1 . 0 3 . 0
3 2 1
-
2 2 1
2 2 1
1 2 1
> +
>
> +
>
+
x x
x x
x x
x x
0 , 0
2 1
> > x x
2 1
5 . 0 4 . 0 x x +
Max
s.t.
. 0 , 0 , 0 , 0
5 . 0 4 . 0 ) ( 5 . 0 1 . 0
4 . 0 6 . 0 ) ( 5 . 0 3 . 0
6 ) ( 6 7 . 2
3 2 2 1
3 2 2 1
3 2 2 1
3 2 2 1
> > > >
s + +
s + +
+ +
+
+
+
+
y y y y
y y y y
y y y y
y y y y
Max
s.t.
. 0 , URS : , 0
5 . 0 4 . 0 5 . 0 1 . 0
4 . 0 6 . 0 5 . 0 3 . 0
6 6 7 . 2
3 2 1
3 2 1
3 2 1
3 2 1
> >
s + +
s + +
+ +
y y y
y y y
y y y
y y y
Application of
Complementary Slackness Conditions.
Example: Solving a problem with 2 functional
constraints by graphical method.
0 , ,
10 4
30 4 3 . .
3 7 2 max
3 2 1
3 2 1
3 2 1
3 2 1
>
s +
s + +
+ =
x x x
x x x
x x x t s
x x x Z
Optimal solution
x
1
=10
x
2
=0
x
3
=0