1 / 25

LP-Based Branch-and-Bound

LP-Based Branch-and-Bound. Example 1. Rounding the LP-Relaxation. Step 1: Solve the LP Relaxation. Optimal LP Solution: x = 2.25, y = 3.75, z = 41.25. x + y= 6. (2.25, 3.75). 5x + 9y = 45. Feasible Region. (0,5). (0,0). (6,0). Step 2: Branch on x.

edan-mccoy
Download Presentation

LP-Based Branch-and-Bound

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. LP-Based Branch-and-Bound

  2. Example 1

  3. Rounding the LP-Relaxation

  4. Step 1: Solve the LP Relaxation Optimal LP Solution: x = 2.25, y = 3.75, z = 41.25

  5. x + y= 6 (2.25, 3.75) 5x + 9y = 45 Feasible Region (0,5) (0,0) (6,0)

  6. Step 2: Branch on x • Optimal LP Solution: x = 2.25, y = 3.75 z = 41.25 • Create two subproblems • LP1 : add the constraint x  2 to LP0 • LP2 : add the constraint x  3 to LP0 The Branch-and-Bound Tree LP0 z = 41.25 x = 2.25, y = 3.75 x  2 x  3 LP1 LP2

  7. x = 3 (2.25, 3.75) LP 1 LP 2 x = 2 Feasible Regions x + y= 6 (0,5) (0,0) (6,0) 5x + 9y = 45

  8. Bounding Step: Solve Subproblem 1 Optimal LP1 Solution: x = 2, y = 3.89, z = 41.11

  9. Bounding Step: Solve Subproblem 2 Optimal LP2 Solution: x = 3, y = 3, z = 39

  10. The Branch-and-Bound Tree LP0 z = 41.25 x = 2.25, y = 3.75 x  2 x  3 LP1 z = 41.11 x = 2, y = 3.89 LP2 z = 39 x=3, y=3 Step 3: Fathoming Step • Subproblem 2 has an integer-valued solution. • The solution x = 3, y = 3 becomes the incumbent solution with z* = 39 and zup = 41 • Fathom the node for Subproblem 2 • Observation z*/zup = 39/41 = 95%

  11. x = 3 x = 2 Feasible Regions Optimal LP solution in region 1: (2, 3.89) x + y= 6 (0,5) Optimal IP solution in region 2. (0,0) (6,0) 5x + 9y = 45

  12. Generic Branching Step • Pick an unfathomed node and branch on an integer variable xi that has a non-integer value  in the optimal solution to that node’s subproblem. • Create two subproblems • One with the constraint xi  • One with the constraint xi 

  13. Branching Step: Branch on y The Branch-and-Bound Tree LP0 z = 41.25 x = 2.25, y = 3.75 • Optimal LP1 Solution: x = 2, y = 3.89 z = 41.11 • Create two subproblems • LP3 : add the constraint y  3 to LP1 • LP4 : add the constraint y  4 to LP1 x  2 x  3 LP1 z = 41.11 x = 2, y = 3.89 LP2 z = 39 x=3, y=3 y  3 y  4

  14. LP 4 y = 4 y = 3 LP 3 Feasible Regions x = 3 x + y= 6 (0,5) (0,0) (6,0) 5x + 9y = 45 x = 2

  15. Bounding Step: Solve Subproblem 3 Optimal LP3 Solution: x = 2, y = 3, z = 34

  16. Bounding Step: Solve Subproblem 4 Optimal LP4 Solution: x = 1.8, y = 4, z = 41

  17. Fathoming Step LP0 z = 41.25 x = 2.25, y = 3.75 x  3 x  2 LP2 z=39 x=3, y=3 LP1,z = 41.11, x = 2, y = 3.89 y  4 y  3 LP3,z = 34, x = 2, y = 3 LP4,z = 41 x = 1.8, y = 4 • Subproblem 3 has an integer-valued solution. • Since 34 is less than 39, don’t update z*. • Fathom the node for Subproblem 3. • Branching Step: branch on x in subproblem 4

  18. Feasible Regions x = 1 x = 3 x + y= 6 (0,5) y = 4 y = 3 LP 5 (0,0) (6,0) 5x + 9y = 45 x = 2

  19. Bounding Step: Solve Subproblem 5 Optimal LP5 Solution: x = 1, y = 4.44, z = 40.56

  20. Bounding Step: Solve Subproblem 6 LP6 is infeasible.

  21. Fathoming Step LP0 z = 41.25 x = 2.25, y = 3.75 x  3 x  2 LP2 z=39 x=3, y=3 LP1,z = 41.11, x = 2, y = 3.89 y  4 y  3 Since the largest z value for any unfathomed node is 40.56, we can update zup to 40. LP3,z = 34, x = 2, y = 3 LP4,z = 41 x = 1.8, y = 4 x  2 x  1 LP5,z = 40.56 x = 1, y = 4.44 LP6 Infeasible

  22. Branching Step • We have an incumbent solution with z*=39 and upper bound of zup = 40. Our incumbent solution is within 39/40 = 97.5% of optimality (i.e., the optimality gap is 2.5%). • Branch on y in LP5 • LP7: add constraint y  4 to LP5 • LP8: add constraint y  5 to LP5

  23. Bounding Step: Solve Subproblem 7 Optimal LP7 Solution: x = 1, y = 4, z = 37

  24. Bounding Step: Solve Subproblem 8 Optimal LP8 Solution: x = 0, y = 5, z = 40

  25. Fathoming Step LP0 z = 41.25 x = 2.25, y = 3.75 x  3 x  2 LP2 z=39 x=3, y=3 LP1,z = 41.11, x = 2, y = 3.89 y  4 y  3 LP4,z = 41 x = 1.8, y = 4 LP3,z = 34, x = 2, y = 3 x  2 x  1 LP6 Infeasible LP5,z = 40.56 x = 1, y = 4.44 y  4 y  5 z* = 40 optimal solution: x = 0, y =5 LP7,z = 37 x = 1, y = 4 LP8,z = 40 x =0, y = 5

More Related