GENERAL

FORM OF LPP
•  n variables (instruments), m constraints

Maximization program
Max co = c1 x1 + c2 x2 + .... + cn xn
subject to
a11 x1 + a12 x2 + .... + a1n xn ≤ b1
a21 x1 + a22 x2 + .... + a2n xn ≤ b2
.
.
am1 x1 + am2 x2 + ....+ amn xn ≤ bm
xi ≥ 0 ( i = 1, ..., n)
Minimization program:

Min co = c1 x1 + c2 x2 + .... + cn xn
subject to
a11 x1 + a12 x2 + .... + a1n xn ≥ b1
a21 x1 + a22 x2 + .... + a2n xn ≥ b2
.
.
am1 x1 + am2 x2 + ....+ amn xn ≥ bm
xi ≥ 0 ( i = 1, ..., n)
•  Can convert inequality constraints into equality
constraints by adding slack/surplus variables.

In max program,
a11 x1 + a12 x2 + .... + a1n xn ≤ b1

is replaced by
a11 x1 + a12 x2 + .... + a1n xn + s1 = b1
where s1 is the slack variable.

In min program,
a11 x1 + a12 x2 + .... + a1n xn ≥ b1
is replaced by

a11 x1 + a12 x2 + .... + a1n xn - t1 = b1

where t1 is the surplus variable.
 Thm. (Dantzig) If the LP problem has an optimal solution,
then the optimal solution is an extreme point in the feasible
region.

Possible Outcomes:

a) No feasible solution
2x1 + x2 ≤ 6
5x1 + 4x2 ≥ 40

b) If the feasible region is not bounded from above, LP

problem has a solution if it involves minimization, but none
if it involves maximization.
2x1 + x2 ≥ 6
5x1 + 4x2 ≥ 40
c) Bounding face solution : Degenerate problem
min C = 5w + 5r
subject to
w + 2r ≥ 1
2w + r ≥ 1
1.25 w + 1.25 r ≥ 1
w, r ≥ 0
r

w
 Duality in Linear Programming

The pair of LPPs:

Minimize c1 x1+ c2 x2 + " + cn xn Maximize b1 y1+ b2 y2 +!+ bm ym

subject to subject to
a11 x1 + a12 x2 + … + a1n xn ≥ b1 a11 y1 + a21 y2 +…+ am1 ym ≤ c1
a21 x1 + a22 x2 + … + a2 n xn ≥ b2 a12 y1 + a22 y2 +…+ am2 ym ≤ c2
! "
am1 x1 + am 2 x2 + … + amn xn ≥ bm a1n y1 + a2 m y2 +…+ amm ym ≤ cn
x1 ≥ 0, … , xn ≥ 0 y1 ≥ 0, …, ym ≥ 0

are dual to each other.

The original problem, called the primal problem, is
transformed into a dual problem, by the following
procedures:
a) Change “maximize” to “minimize” and vice-versa.
b) Except for the nonnegativity restrictions, the inequality signs
in the primal constraints are reversed in the dual constraints.
c) The coefficient matrix of the dual constraints is the
transpose of the coefficient matrix of the primal constraints.
d) The row vector of coefficients in the primal objective
function is the column vector of constants in the dual
constraints.
Examples:

a) Max 45x1 + 55x2

subj. to
6x1 + 4x2 ≤ 120
3x1 + 10 x2 ≤ 180
x1, x2 ≥ 0

The dual is given by:

Min 120 y1 + 180 y2
subj. to
6y1 + 3y2 ≥ 45
4y1 + 10y2 ≥ 55
y1, y2 ≥ 0
b) Min 2x1 + 10x2
subj. to
2x1 + x2 ≤ 6
5x1 + 4x2 ≥ 20
x1, x2 ≥ 0
Must be written in general form:
Min 2x1 + 10x2
subj. to
-2x1 - x2 ≥ -6
5x1 + 4x2 ≥ 20
x1, x2 ≥ 0
The dual is
Max -6y1 + 20y2
subj. to
-2y1 + 5y2 ≤ 2
-y1 + 4y2 ≤ 10
y1, y2 ≥ 0
Thm. If x is primal feasible and y is dual feasible then
VOF(x) ≥ VOF (y), i.e., the minimand can never fall below the
maximand.

Thm. If x* is primal feasible and y* is dual feasible, and

VOF (x*) = VOF(y*), then x* and y* are optimal solutions.

Thm. If the primal is unbounded, then the dual is

infeasible.