Professional Documents
Culture Documents
Feng Chen
Mar 18 2010 1
Copyright Feng Chen 2004-2010. All rights reserved.
Enumerate Algorithm
min z cT x
s.t. Ax b
x0
Convert methods
<= , Slack variables
p >=, excess variables
p Free variable, x=x1-x2
Standard LP with x1 and x2 to be
nonnegative.
Linear
Programming ? For standard LP, can we
get the extreme point by
algebraic calculations.
Converted
Case 2: m>n Such case seems easy, if we can have a method to determine its
infeasibility. (overdetermined)
x1 x2 3
x1 2 x2 6 We like case 1 & 2, it seems easy to deal with.
x1 4 x2 5
Case 3: m<n The case seems hard to get optimal, we do not like, because
x1 x2 3 they give me many solutions, and we need to find the best
x 2 x3 1 one(undetermined).
This case will lead to many feasible solutions. So, we will address
such difficulty case.
Mar 18 2010 Page 7
Copyright Feng Chen 2004-2010. All rights reserved.
Basic solution
In case m<n
x1 x2 3
x2 x3 1
How to find all feasible solutions? Because the number of variable
is larger than the number of equation, we would like to fix some
variables, and solve a normal equation system with left variables.
min z 4 x1 3 x2
Basic Nonbasic Basic
s.t. x1 x2 y3 40 variables Variables Feasible
2 x1 x2 y 4 60 Solution
x1 , x2 , y3 , y 4 0 x1, x2 y3,y4 (20,20,0,0)
Proof.
Given a basic feasible solution
x=(x1,x2,,xn-m, 0,0,,0)
y x
z Contradiction
Mar 18 2010 17
Copyright Feng Chen 2004-2010. All rights reserved.