The Simplex Method

Firstly, the definition of linear programming is an area of linear algebra in which the goal is to maximize or minimize a linear function on a region of variables

whose boundary is defined by linear inequalities and equations. In this variables" means that has the form:

context, "linear function of

The region of

is a convex polytope. If all the vertices of

are known, then the maximum

will occur at one of these vertices. In two dimensions a convex polytope is a

region that is the intersection of a finite set of half-planes (the general idea of a convex polygon). In three dimensions, a convex polytope is solid region that is the intersection of a finite set of half-spaces (the generalized idea of a convex polyhedron). The generalization in n dimensions is called a polytope. The solution can be constructed using the simplex method and is invented and attributed by George Dantzig in 1947. George Dantzig (1914 ) who was born in Portland, Oregon, tests adjacent vertices of the feasible set (which is a polytope) in sequence so that at each new vertex the objective function improves or is unchanged. The simplex method is very efficient in practice, generally taking to iterations at most (where is the

number of equality constraints), and converging in expected polynomial time for certain distributions of random inputs (Nocedal and Wright 1999, Forsgren 2002). However, its worst-case complexity is exponential, as can be demonstrated with carefully constructed examples (Klee and Minty 1972). The simplex method starts at the origin and follows a path along the edges of the polytope to the vertex where the maximum occurs. The history of the development of the simplex method has been summarized in the article: An Interview with George B. Dantzig: The Father of Linear Programming by Donald J. Albers; Constance Reid; in The College Mathematics Journal, Vol. 17, No. 4. (Sep., 1986), pp. 292-314, Jstor. A different type of methods for linear programming problems are interior point methods, whose complexity is polynomial for both average and worst case. These methods construct a sequence of strictly feasible points (i.e., lying in the interior of the polytope but never on its boundary) that converges to the solution. Research on interior point methods was spurred by a paper from Karmarkar (1984). In practice, one of the best interior-point

methods is the predictor-corrector method of Mehrotra (1992), which is competitive with the simplex method, particularly for large-scale problems. Dantzig's simplex method should not be confused with the downhill simplex method (Spendley 1962, Nelder and Mead 1965, Press et al. 1992). The latter method solves an unconstrained minimization problem in dimensions by maintaining at each iteration

points that define a simplex. At each iteration, this simplex is updated by applying certain transformations to it so that it "rolls downhill" until it finds a minimum. Standard Form of the Linear Programming Problem The standard form of the linear programming problem is to maximize . (1) Maximize of variables

where

for

. where

(2) Subject to the m constraints for . for .

(3) With the primary constraints The coefficients , but the cases and or

can be any real number. It is often the case that can occur.

Setting up the Extended Simplex Tableau For computational purposes, we construct a tableau. The first coefficients matrix , the identity matrix rows consist of the . In the , which are

and the column vector

-row of the tableau the first

elements are the coefficients

called the coefficients of the augmented objective equation. The remainder of the bottom row is filled in with zeros. An extra column on the right will be used in the decision process in solving for the variables.

Column

The non-negative variable linear combination to change an inequality

is called a slack variable and is the amount that the falls short of the bound to an equation, i.e. . It's purpose is we have

(4)

for rows

in the tableau.

The goal of the simplex method is to exchange some of the columns of 1's and 0's of the slack variables into columns of 1's and 0's of the decision variables. Algorithm (Simplex Method). To find the minimum of polytope (i) . and form a system of equations over the convex

Use non-negative slack variables and the initial tableau.

(ii) Determine

the

exchange

variable,

the

pivot

row

and where

pivotal

element.

The exchange variable negative coefficient.

is chosen in the pivot column

is the smallest

The pivot row with

is chosen where the minimum ratio .

occurs for all rows

. The pivot element is

(iii) Perform row operations to zero out elements in the pivotal column the pivot row .

above and below

(iv) Repeat steps (ii) and (iii) until there are no negative coefficients row.

in the bottom