Approach to solve LP

Definitions
Adjacent Extreme Points: Two extreme points of a polygon are called adjacent extreme points if the line segment joining them is an edge. Local Optimal Solution: If a solution is optimal in its neighbourhood. Global Optimal Solution: If it is optimal globally, i.e. with respect to all possible feasible solutions Key Result: For an LP local solution is the global solution. If a corner point of the feasible region yields the objective value better than the neighbouring extreme points then it is the optimal point. (how??)

Conversion of linear inequalities to equalities

x1 2 x1 - s1 =2 s1( 0) is a surplus variable.

x1 + x2 8 x1 + x2 + s2 =8 s2( 0) is a slack variable.

Convert all the constraints to equality by introducing slack/surplus variables. Why we do so? It is difficult to deal with inequations rather than equations. Studying a solution of inequations means studying a solution of equations in higher dimension.

Basic Feasible Solution (BFS)

set (n-m) variables to any arbitrary values (say zero) leaving m equations in m unknowns yielding a unique solution. Such a solution is called a basic solution.
(n-m) variables set to zero are called non-basic variables.

Remaining m variables with (usually) non-zero values are called basic variables.

Note: Two adjacent corner points differ in only one basic (or non-basic) variable.

Example
3x1+2x2+5x3 = 16 4x1+x2+5x3 = 18
Let x3 = 0 (non-basic variable)
3x1+2x2 = 16 4x1+x2 = 18 Solution: x1 = 4, x2 = 2 Basic solution : (4,2,0) Let x1=0 (non-basic variable)

2x2+5x3 = 16
x2+5x3 = 18 Solution: x3=4, x2 = -2 Basic solution : (0,-2, 4)

Example
Original Variables : 2 Inequality Constraints : 3 (m)
3 2 x2 III 1 x1 I II 4

Slack Variables Added : 3

Total Variables : 2 + 3 = 5 (n) X vector is ( x1, x2, s1, s2, s3 )

Example
Corner point 1 : ( 0,0, s1, s2, s3 ) Corner point 2 : ( 0, x2, 0, s2, s3 ) Corner point 3 : ( x1, x2, 0, 0, s3 )
x2 III 1 x1 2 I 3 II 4

Corner point 4 :
( x1, x2, s1, 0, 0 ) Corner point 5 : ( x1, 0, s1, s2, 0 )
5

In each case exactly two i.e. n-m variables are at value zero. These two are the non-basic variables, and the other three are basic variables.

Check if the given solution is a BFS

BFS:
Must be a solution i.e. must satisfy the equations.

At least (n m) variables have to be zero (At the most m

variables can be non-zero).

columns corresponding to the remaining m (non-zero) variables must be LI.

Examples

BFS (Summary)
XB = B1b is called the basic solution of AX = b.

B is called a basis, as basis changes the solutions changes.

Columns of A which form basis are called basic vectors. Basic variables are the variables associated with basic vectors

Non-basic variables are always zero.

BFS
In a BFS
Non-basic variables are zero. (always)

Basic variables are Non-negative.

Degenerate BFS:
A BFS with at least one of the basic variable equal to zero is known as a degenerate BFS otherwise it is called nondegenerate.

How many BFS are possible for a system AX = b, X 0 with m equations in n unknowns???
At the most nCm (how?)