This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: MthSc 810: Mathematical Programming Lecture 7 Pietro Belotti Dept. of Mathematical Sciences Clemson University September 15, 2011 Reading for today: textbook, Sections 3.1 and 3.2 Reading for Sep. 20: Sections 3.3 and 3.4 Homework #4 out today – due September 22. Recap Consider the Linear Optimization problem LP : min { c ⊤ x : A x = b , x ≥ } . If ◮ P = { x ∈ R n + : A x = b } contains ≥ 1 extreme points, and ◮ LP admits an optimal solution, then there exists an optimal solution that is an extreme point. Even more: if P has at least one extreme point, then ◮ either there exists an optimal that’s an extreme point, or ◮ the cost is −∞ (i.e., the problem is unbounded) The standard form is convenient min { c ⊤ x : A x = b , x ≥ } . ◮ There exists an extreme point since P ⊆ R n + . ◮ If A ∈ R m × n , then rank ( A ) = m ≤ n . ◮ (If rank ( A ) < m , we can eliminate some equations) ◮ P is the intersection of a subspace ( A x = b ) with a cone (the first orthant R n + ) Basic solutions ◮ Select m linearly independent columns (variables) of A ◮ Call B this submatrix and denote A = ( B  N ) ◮ The variable vector x can be partitioned accordingly as x = x B x N ( x B ∈ R m , x N ∈ R n − m ) ◮ So A x = b is equivalent to B x B + N x N = b ◮ Then x B = B − 1 b − B − 1 N x N ⇒ x = x B x N = B − 1 b is a basic solution ◮ Why? there are n active constraints: A x = b and x N ≥ , or B N I x B x N = b If, in addition, B − 1 b ≥ 0, we have a BFS.0, we have a BFS....
View
Full
Document
This note was uploaded on 03/14/2012 for the course MTHSC 810 taught by Professor Staff during the Fall '08 term at Clemson.
 Fall '08
 Staff
 Math

Click to edit the document details