Lecture 10 - Solving Integer Programming Branch and Bound...

Info iconThis preview shows pages 1–9. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Solving Integer Programming Branch and Bound Method Dynamic Programming
Background image of page 1

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

View Full DocumentRight Arrow Icon
2 IP and Its LP Relaxation Given an IP, we can obtain an LP by removing the integrality constraints Referred to as the LP relaxation Upper bound for a maximization IP problem – Z IP = optimal objective function value to the IP – Z LP = optimal objective function value to the LP relaxation – We have Z IP ≤ Z LP Because the feasible region of LP relaxation contains the feasible region of IP – Z LP is an upper bound of the IP
Background image of page 2
3 LP Relaxation Maximize 4 x 1 +9 x 2 +6 x 3 Subject to 5 x 1 +8 x 2 +6 x 3 ≤12 x 1 , x 2 , x 3 are binary variables For the LP relaxation (0≤ x i ≤1), we have x 1 =0, x 2 =1, x 3 =2/3, and Z LP =13 Thus, the optimal solution to the IP 13 – The actual Z IP =10
Background image of page 3

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

View Full DocumentRight Arrow Icon
4 Branch An IP may be gradually decomposed into a series of smaller IP problems • Example: x 1 is a binary variable, then we may have two smaller problems, IP1 and IP2: – Let Z IP1 be the optimal solution of IP1 with x 1 =0 – Let Z IP2 be the optimal solution of IP2 with x 1 =1 – Z IP = max(Z IP1 , Z IP2 )
Background image of page 4
5 Bound • Suppose we already have Z IP1 for IP. Consider IP2: – Solving the LP Relaxation of IP2 to obtain Z LP2 – If Z LP2 ≤ Z IP1 , we do not need to solve IP2, since Z IP2 ≤ Z LP2 ≤ Z IP1 , so Z IP1 must be the optimal solution to the original IP This is called BOUND For each smaller IP, we can solve its LP relaxation and see if we can fathom it directly, meaning that we will no longer need to solve it
Background image of page 5

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

View Full DocumentRight Arrow Icon
6 The Branch and Bound Tree k -1 k k+ 2 k +1 x k -1 =1 x k -1 =0 x k =1 x k = 0 x k+ 1 =1 x k +1 =0 x k +2 =1 x k +2 =0 k+ 3 k+ 4 k+ 5 k+ 6 k+ 7 A partial tree
Background image of page 6
7 The Branch and Bound Tree k -1 k k+ 2 k +1 x k -1 =3 x k -1 = 2 x k =7 x k = 6 x k+ 1 =2 x k +1 = 1 x k +2 =3 x k +2 =2 k+ 3 k+ 4 k+ 5 k+ 6 k+ 7 A partial tree x k- 1 =3.6 in the current LP Solution
Background image of page 7

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

View Full DocumentRight Arrow Icon
8 General Ideas Suppose we have a feasible solution of the IP at hand, the objective function value is Z B Z B is a lower bound of the problem Let Z k be objective function value of the LP relaxation at some node k . Whenever an IP
Background image of page 8
Image of page 9
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 01/05/2011 for the course IELM IELM202 taught by Professor D during the Fall '10 term at HKUST.

Page1 / 26

Lecture 10 - Solving Integer Programming Branch and Bound...

This preview shows document pages 1 - 9. Sign up to view the full document.

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