Chapter 3 - MA3252 LINEAR AND NETWORK OPTIMIZATION Topic 3:...

Info iconThis preview shows pages 1–13. Sign up to view the full content.

View Full Document Right Arrow Icon
MA3252 LINEAR AND NETWORK OPTIMIZATION Topic 3: The Simplex Method 1 / 81
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Consider the standard form of a LP min c 0 x s.t. Ax = b x 0 . Assume A is an m × n matrix and rank ( A ) = m with m n . Let P = { x R n | Ax = b , x 0 } . If P 6 = φ , then P has an extreme point since it does not contain a line. Therefore, either the optimal value is unbounded or there exists an optimal solution which can be found among the finite set of extreme points. 2 / 81
Background image of page 2
We now develop a procedure to identify the optimal solution. We derive conditions to check for optimality. Based on a procedure for obtaining a new basic feasible solution from an old basic feasible solution and the optimality conditions, the simplex method will be developed. 3 / 81
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Optimality Conditions In this section, we derive optimality conditions to check whether a basic feasible solution is optimal. This is useful in the development of simplex method. The optimality conditions also provide a clue for identifying a direction to improve the objective value in a neighbourhood of a basic feasible solution. 4 / 81
Background image of page 4
Let x be a basic feasible solution with the set { B (1) , ··· , B ( m ) } of basic indices, so that B = ( A B (1) A B (2) A B ( m ) ) and x B = x B (1) · · · x B ( m ) = B - 1 b 0 . 5 / 81
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Definition For a polyhedron P , and a point x P , a vector d is a feasible direction if x + θ d P for some θ > 0. Consider a BFS x = ( x B , x N ) 0 and a feasible direction d = ( d B , d N ) such that the direction brings non-basic variable j into the basis, leaving all other non-basic variables at zero. 6 / 81
Background image of page 6
Then the j-th basic direction is defined as: d j = 1 and d i = 0 for non-basic variables i 6 = j d B = - B - 1 A j To guarantee the direction is feasible, we need B - 1 b - θ B - 1 A j 0 . 7 / 81
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
How far we can move in this direction d , depends on the BFS x ? 1 If B - 1 A j 0 , then the polyhedron is unbounded in the x j -direction. Eg: P = { ( x 1 , x 2 ) R 2 | x 1 = 1 , x 1 , x 2 0 } . 8 / 81
Background image of page 8
2 If x is nondegenerate, i.e. x B > 0 . If ( B - 1 A j ) k > 0 for some component k , then θ ( B - 1 b ) k ( B - 1 A j ) k . Ratio Test Let θ * = min ± ( B - 1 b ) k ( B - 1 A j ) k ² ² ² ( B - 1 A j ) k > 0 ³ = ( B - 1 b ) l ( B - 1 A j ) l . The feasible solution ˆ x = x + θ * d is a basic feasible solution which is adjacent to x , with associated basic variables { x B (1) , ··· , x B ( l - 1) , x B ( l +1) , , x B ( m ) , x j } . Eg: P = { ( x 1 , x 2 ) R 2 | x 1 + x 2 = 1 , x 1 , x 2 0 } . 9 / 81
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
3 If x is degenerate, i.e. there is some zero basic variable x B ( k ) , then the nonzero vector d may not be a feasible direction. This happens if a basic variable x B ( k ) = 0 and d B ( k ) < 0, so that x + θ d 6∈ P for all positive scalars θ . Hence θ * = 0. Eg: P = { ( x 1 , x 2 , x 2 , x 3 ) R 4 | x 1 + x 2 + x 3 = 1 , x 1 + x 4 = 1 , x 1 , x 2 , x 3 , x 4 0 } . 10 / 81
Background image of page 10
11 / 81
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Moving from x to ˆ x = x + θ d , the change in objective function value is c 0 ˆ x - c 0 x = θ c 0 d . Suppose d , with d B = - B - 1 A j , is the feasible direction obtained as before. Then the change in the objective value for a unit increase in x j is c 0 d = c j - c 0 B B - 1 A j , where c B = ( c B (1) , c B (2) , ··· , c B ( m ) ) 0 .
Background image of page 12
Image of page 13
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 12/02/2011 for the course MATH 3252 taught by Professor Tanbanpin during the Spring '10 term at National University of Singapore.

Page1 / 81

Chapter 3 - MA3252 LINEAR AND NETWORK OPTIMIZATION Topic 3:...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online