Scribd
Upload
66 views17 pages

Lecture 3 The Simplex Method

The document outlines the simplex method for solving linear programs, including how it works by moving between basic feasible solutions until an optimal solution is found using an optimality test. It also discusses using the augmented matrix form and finding an initial basic feasible solution, either through a phase 1 approach or big M method, before the simplex method iterations can begin.

Uploaded by

Vishruti Desai
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
66 views17 pages

Lecture 3 The Simplex Method

The document outlines the simplex method for solving linear programs, including how it works by moving between basic feasible solutions until an optimal solution is found using an optimality test. It also discusses using the augmented matrix form and finding an initial basic feasible solution, either through a phase 1 approach or big M method, before the simplex method iterations can begin.

Uploaded by

Vishruti Desai
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 17

ISM 206

Lecture 3
The Simplex Method
Announcements
Outline

• LP so far
• Why we can look only at basic feasible solutions
• Optimality conditions
• The simplex method
• The step from one bfs to the next
• Tableu method
• Phase I: Finding an initial BFS
LP so far
• Formulated LP’s in various contexts
• Transform any LP into a standard form LP
• Intuition of simplex method: Find the best corner
point feasible solution
• Math required:
– Corner point Feasible or basic feasible solutions
correspond to a set of n active constraints
– Any set of active constraints corresponds to a basis
from the matrix A
– The basis is a set of linearly independent columns
Standard Form
max c1 x1  c2 x2  ...  c N x N
subject to
a11 x1  a12 x2  ...  a1N x N  b1
a21 x1  a22 x2  ...  a2 N x N  b2
...
aM 1 x1  aM 2 x2  ...  aMN x N  bM
x j  0, j  1..N

max cx
Concise version: subject to Ax  b
x0
A is an m by n matrix: n variables, m constraints
Standard Form to Augmented Form
max c1 x1  c2 x2  ...  c N x N max c1 x1  c2 x2  ...  c N x N
subject to subject to

a11 x1  a12 x2  ...  a1N x N  b1 a11 x1  a12 x2  ...  a1N x N  x1  b1

a21 x1  a22 x2  ...  a2 N x N  b2 a21 x1  a22 x2  ...  a2 N x N  x2  b2
... ...

aM 1 x1  aM 2 x2  ...  aMN x N  bM aM 1 x1  aM 2 x2  ...  aMN x N  x N  bM

x j  0, j  1..N x j , xi  0, j  1..N , i  1..m

max cx max cx


 x
subject to Ax  b subject to[ A, I ]    b
x
x0 x0
A is an m by n matrix: n variables, m constraints
Solutions, Extreme points and
bases
• Key fact:
– If a LP has an optimal solution, then it has an optimal
extreme point solution (proved today)
• Basic Feasible solution (Corner Point Feasible):
– The vector x is an extreme point of the solution space
iff it is a bfs of Ax=b, x>=0
• If A is of full rank then there is at least one basis
B of A
– B is set of linearly independent columns of A
• B gives us a basic solution
– If this is feasible then it is called a basic feasible
solution (bfs) or corner point feasible (cpf)
Optimal basis theorem
Theorem
If a LP in standard form has a finite optimal
solution then it has an optimal basic
feasible solution
Proof
Requires the representation theorem…
Simplex Method
• Checks the corner points
• Gets better solution at each iteration

1. Find a starting solution


2. Test for optimality
– If optimal then stop
3. Perform one iteration to new CPF (BFS)
solution. Back to 2.
Simplex Method: basis change
• One basic variable is replaced by another
• The optimality test identifies a non-basic
variable to enter the basis
– The entering variable is increased until one of
the other basic variables becomes zero
– This is found using the minimum ratio test
– That variable departs the basis
Idea of simplex iterations
• New matrix A is an m by n matrix
– m constraints
– n variables
• Z = objective value (another variable)
• m+1 constraints, n variables
– We can rearrange m+1 equations, trying to
• Maximize Z
• Keeping x >=0
• Consider basic solutions
– m x-variables are nonzero
– All others are zero
The simplex method
• Example
• Table version
Questions and Break
A basic feasible solution
• B=basis of A. max c' x
s.t. Ax  b
• Write LP in terms of basis
x0
A  [B N ]
Ax  BxB  Nx N  b

xB   m , x N   n  m
x N  0, xB  B 1b

X is a basic solution of the LP


X is a basic feasible solution if it is feasible!
(example)
Optimality of a basis
We want to test of a basic feasible solution is optimal max cB ' xB  c N ' x N
Use the basic notation from before s.t.
BxB  Nx N  b
BxB  Nx N  b xB , x N  0
xB  B 1 (b  Nx N )

c' x  cB ' xB  c N ' x N


 cB ' B 1 (b  Nx N )  cN ' xN
 cB ' B 1b  cB ' B 1 Nx N  cN ' x N
 z0   ( z j  c j )x j
j

z0  cB ' B 1b
z j  cB ' B 1 A. j j  N
Optimality test

max c' x  z0   ( z j  c j )x j
j

Test :
z j  c j  0 j  N
Finding an initial bfs
• The ‘phase 1’ approach
• The ‘big M’ method

You might also like