Lecture17 - Integer Programming IE418 Lecture 17 Dr. Ted...

Info iconThis preview shows pages 1–7. Sign up to view the full content.

View Full Document Right Arrow Icon
Integer Programming IE418 Lecture 17 Dr. Ted Ralphs
Background image of page 1

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

View Full DocumentRight Arrow Icon
IE418 Lecture 17 1 Reading for This Lecture Nemhauser and Wolsey Sections II.1.1-II.1.3, II.1.6 Wolsey Chapter 8 Valid Inequalities for Mixed Integer Linear Programs, G. Cornuejols (2006)
Background image of page 2
IE418 Lecture 17 2 Describing conv( S ) We now switch to the case of a pure integer program with explicit nonnegativity constraints (the reason will become clear). z IP = max { cx | x S } , S = { x Z n + | Ax b } . We have just seen that in theory, it would be possible to generate a complete description of conv( S ) . So why aren’t IPs easy to solve? The number of inequalities is generally HUGE ! The number of facets of the TSP polytope for an instance with 120 nodes is more than 10 100 times the number of atoms in the universe . It is physically impossible to write down a description of this polytope. Not only that, but it is very difficult in general to generate these facets (this problem is not in P in general).
Background image of page 3

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

View Full DocumentRight Arrow Icon
IE418 Lecture 17 3 For Example For a TSP of size 15 The number of subtour elimination constraints is 16,368 . The number of comb inequalities is 1 , 993 , 711 , 339 , 620 . These are only two of the know classes of facets for the TSP. For a TSP of size 120 The number of subtour elimination constraints is 0 . 6 × 10 36 ! The number of comb inequalities is approximately 2 × 10 179 !
Background image of page 4
IE418 Lecture 17 4 Basic Bounding Methods Our discussions of branch and bound have so far focused on the use of three basic bounding methods. LP relaxation Lagrangian relaxation Dantzig-Wolfe decomposition Recall from Lecture 13 that the bound produced by Lagrangian relaxation and Dantzig-Wolfe decomposition is z D = max { cx | A 1 x b 1 , x conv( S LR ) } , which is an improvement over that produced by solving the LP relaxation. Producing the bound z D depends on our ability to efficiently optimize over conv( S LR ) . Can we improve the LP relaxation in some way?
Background image of page 5

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

View Full DocumentRight Arrow Icon
IE418 Lecture 17 5 Cutting Planes Because of the equivalence of optimization and separation, we could also produce the bound z D by dynamic generation of valid inequalities. Recall that the inequality denoted by ( π, π 0 ) is valid for a polyhedron P if πx π 0 x ∈ P . The term cutting plane usually refers to an inequality valid for conv( S ) , but which is violated by the solution obtained by solving the (current) LP relaxation. Note that this is not a very precise definition and the term is a bit colloquial, but we will use it anyway. Cutting plane methods attempt to improve the bound produced by the LP
Background image of page 6
Image of page 7
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 08/06/2008 for the course IE 418 taught by Professor Ralphs during the Spring '08 term at Lehigh University .

Page1 / 20

Lecture17 - Integer Programming IE418 Lecture 17 Dr. Ted...

This preview shows document pages 1 - 7. Sign up to view the full document.

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