|
applet-magic.com Thayer Watkins Silicon Valley & Tornado Alley USA |
|---|
|
Problems in Linear Programming |
Let X be a column vector of outputs and P a row vector of the outputs'
prices. R is a column vector of the amounts of resources available and
A is matrix which gives the resources use per unit of outputs. The primal
problem is to:
maximize V = PX
subject to:
AX < R
Dual problem:
If slack variables with zero prices are added then the primal is:
The slack variables have zero prices so P' is the price vector augmented with a zero for each slack variable.
A basis is an output vector in which the number of nonzero outputs, including slack variables, is no more than the number of resources, dim(R). For any basis the order of the variables can be changed so the nonzero (or basis) variables are first and the nonbasis variables are last. Such a reordering of the variables also reorders the elements of P' and A'.
Suppose X' is partitioned into basis variables, XB, and nonbasis variables, XNB; i.e.,
| | XB | | ||
| X' | = | | | |
| | XNB | |
Since the price vector is a row vector it is partitioned as P' = [PB PNB]. Likewise the coefficients matrix A can be partitioned as
Then the objective function can be represented as
and the constraints as
If the XNB variables are set to any feasible values then XB variables are determined as
The value of the output is then
When XNB = 0 then V = PBAB,B-1R. For an optimum it must be that the marginal change for any element in XNB is nonpositive; i.e.,
This means that
The marginal values of the resources MRV are given by
It then follows that
Furthermore, the cost of production based upon those marginal values for the elements of XB is
For the nonbasic elements the costs of production are:
|
HOME PAGE OF Thayer Watkins |