This action might not be possible to undo. Are you sure you want to continue?

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

)

Lecture 17 Integer linear programming

• integer linear programming, 0-1 linear programming • a few basic facts • branch-and-bound

17–1

Deﬁnition

integer linear program (ILP) minimize cT x subject to Ax ≤ b, x ∈ Zn

c

Gx = d

mixed integer linear program: only some of the variables are integer 0-1 (Boolean) linear program variables take values 0 or 1

Integer linear programming 17–2

m j=1 xij = 1. . . . . . . n 17–3 Linear programming relaxation the LP obtained by deleting the constraints x ∈ Zn (or x ∈ {0. m. . . . . . . . xij ≤ yj . yj ∈ {0. . . m clients • ci. j = 1.Example: facility location problem • n potential facility locations. . . n: cost of serving client i from location j determine optimal location: minimize subject to n m n j=1 cj yj + i=1 j=1 dij xij n i = 1. . m. . . i = 1. 1} • yj = 1 if location j is selected • xij = 1 if location j serves client i a 0-1 LP Integer linear programming j = 1. . i = 1 . i = 1. then it also solves the integer LP equivalent ILP formulations of the same problem can have diﬀerent relaxations c c Integer linear programming 17–4 . . 1}n) is called the LP relaxation • provides a lower bound on the optimal value of the integer LP • if the solution of the relaxation has integer components. xij . . n: cost of opening a facility at location i • dij .

. . .Strong formulations the convex hull of the feasible set S of an ILP is: K conv S = i=1 λixi xi ∈ S. the solution of the ILP also solves the relaxation minimize cT x subject to x ∈ conv S Integer linear programming 17–5 Branch-and-bound algorithm minimize cT x subject to x ∈ P where P is a ﬁnite set general idea: • decompose in smaller problems minimize cT x subject to x ∈ Pi where Pi ⊂ P. i = 1. i λi = 1 (the smallest polyhedron containing S) c for any c. . λi ≥ 0. K • to solve subproblem: decompose recursively in smaller problems • use lower bounds from LP relaxation to identify subproblems that don’t lead to a solution Integer linear programming 17–6 .

3. and P= x ∈ Z2 + x2 2 1 x1 + x2 ≤ 1.75 +∞ −9.00.00.07) (2.example minimize cT x subject to x ∈ P where c = (−2.14) (3.33 −9. 1. 0.56 −10. 0.00) p⋆ −10.00) (0.00 −9.00 −9. 1.00) (3.33) (2.00) (4. 2) Integer linear programming 17–7 tree of subproblems and results of LP relaxations: P0 x1 ≤ 2 P1 x2 ≤ 2 P3 x2 ≥ 3 P4 x1 = 3 P7 x2 = 0 P9 x1 = 4 P11 x2 ≤ 1 P5 x1 ≥ 4 P8 x2 = 1 P10 x1 ≥ 5 P12 x1 ≥ 3 P2 x2 ≥ 2 P6 P0 P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 x⋆ (2.00.00 −9.00 +∞ Integer linear programming 17–8 .44) (4. 2.00.50. 9 4 −c 1 1 x1 + x2 ≤ 1 7 3 x1 optimal point: (2.00 −10. −3).00. 2.38.17.43 −10. 1.00) (4.00) (3.00. 0. 2.00.00 +∞ −8.

x2 ≤ 2 Integer linear programming 17–9 • P6: the problem minimize cT x subject to x ∈ P. P2. P1..00 • P3: the solution of minimize cT x subject to x ∈ P. 2) x1 ≥ 3 x1 ≤ 2. is greater than or equal to −10. P3.conclusions from subproblems: • P2: the optimal value of minimize cT x subject to x ∈ P. x2 ≥ 2 then after solving subproblem P4 we can conclude that (2. 2) is optimal Integer linear programming 17–10 . x1 ≤ 3. is infeasible suppose we enumerate the subproblems in the order P0 .. . is (2.

mark all nodes in the subtree with root k as inactive ˆ 3. mark all nodes in the subtree with ˆ root k as inactive. i ∈ I3 where I1. xi ∈ {0. then set U := cT x and save x ˆ ˆ ˆ as the best feasible point found so far 4. n} Integer linear programming 17–11 i ∈ I2 branch-and-bound method set U = +∞. i ∈ I1. if moreover cT x < U .branch-and-bound for 0-1 linear program minimize cT x subject to Ax ≤ b. 1}. if cT x ≥ U . . xi = 1. i ∈ I3 let x be the solution of the relaxation ˆ 2. i ∈ I1 k xi = 1. I3 partition {1. select an active node k. go to step 1 Integer linear programming 17–12 . 1}n x1 = 0 x2 = 0 x2 = 0 x2 = 1 can solve by enumerating all 2n possible x. every node represents a problem minimize cT x subject to Ax ≤ b xi = 0. mark all nodes in the tree as active 1. . I2. and solve the corresponding LP relaxation minimize cT x subject to Ax ≤ b k xi = 0. otherwise. x1 = 1 x2 = 1 x ∈ {0. . mark node k as inactive 5. if all components of x are 0 or 1. i ∈ I2 k 0 ≤ xi ≤ 1. .

- WRP Mapping Grid General Mathematics 2012
- v3
- Oaugwebinar2013dimensions 150721175658 Lva1 App6892
- Taylor theorem for two variable
- Life Table
- Lect15_ProbCurrent1
- Proper Graphing
- Optimization
- 09. Merge Sort
- 100ktwins
- JK Analisis
- 4 Channels of Math - No AZtec or Smart Board
- 13_excel_short_cut
- Line Polygon Clipping
- Integration Revision Sheet
- Cap Bud Selection Methods.ppt
- dqTransformation-2
- Applying Superposition
- VHDL PPT
- LD12-01
- Domain Decomposition Strategies With Black Box Subdomain Solver
- Dara Generator
- MCQ Normal,Probability, Confidence
- 22_AntIndex
- ECE 440 Cheat Sheet
- Rec Bookcontoh
- Lecture-03-SecretKeyCryptography_2.pdf
- SAT II Math Level 2 Subject Test Notes
- autocomplete
- 08L9

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd