This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Lecture 17 Linear Programs with Mixed Constraints UCSD Math 171A: Numerical Optimization Philip E. Gill http://ccom.ucsd.edu/~peg/math171 Friday, February 20, 2009 Recap: the phase1 LP We are finding a feasible point for the constraints Ax ≥ b , x ≥ 0. The phase1 LP is the linear program: minimize θ ∈ R , x ∈ R n θ subject to a T 1 x + θ ≥ b 1 a T 2 x + θ ≥ b 2 . . . . . . a T m x + θ ≥ b m θ ≥ UCSD Center for Computational Mathematics Slide 2/28, Friday, February 20, 2009 Recap: the phase1 LP In matrix form, the phase1 LP is: minimize ¯ x ∈ R n +1 ¯ c T ¯ x subject to ¯ A ¯ x ≥ ¯ b , ¯ x ≥ with m general constraints, n + 1 simple bounds n + 1 variables ⇒ m + n + 1 constraints and n + 1 variables UCSD Center for Computational Mathematics Slide 3/28, Friday, February 20, 2009 Solving the phase1 LP Simplex iterates: x θ ! x 1 θ 1 ! ··· x ‘ θ ‘ ! Working sets: ¯ A ¯ A 1 ··· ¯ A ‘ ↑ last The optimal vertex has n + 1 constraints in the working set. If a feasible point exists, then θ ‘ = 0 ⇒ constraint θ ≥ 0 enters the working set at iteration ‘ 1. If no feasible point exists, phase1 will terminate with θ ‘ > ⇒ θ ‘ is the smallest shift that makes the constraints feasible. UCSD Center for Computational Mathematics Slide 4/28, Friday, February 20, 2009 Assume that the constraints are feasible, i.e., θ ‘ = 0. The last working set is ¯ A ‘ ¯ x ‘ = ¯ b ‘ , where ¯ A ‘ includes rows of ( A e ) rows of I n +1 row e T n +1 (just added) Assume that e T n +1 entered as the last row of ¯ A ‘ , then ¯ A ‘ = A ‘ z 1 ! and ¯ b ‘ = b ‘ ! where A ‘ and b ‘ are rows of A and b , and z has elements 0 or 1. UCSD Center for Computational Mathematics Slide 5/28, Friday, February 20, 2009 From the previous slide: ¯ A ‘ = A ‘ z 1 ! and ¯ b ‘ = b ‘ ! If ¯ A ‘ ¯ x ‘ = ¯ b ‘ and ¯ x ‘ ≥ 0 then A ‘ z 1 ! x ‘ θ ‘ ! = b ‘ ! ⇒ A ‘ x ‘ + z θ ‘ = b ‘ θ ‘ = 0 ⇒ A ‘ x ‘ = b ‘ so that A ‘ defines an initial working set for phase 2....
View
Full Document
 Winter '08
 staff
 Math, Operations Research, Linear Programming, Optimization, UCSD Center for Computational Mathematics, UCSD Center, Computational Mathematics

Click to edit the document details