introduction to integer programming modeling and methods

70
Introduction to Integer Programming Modeling and Methods Michael Trick Carnegie Mellon University CPAI-OR School, Le Croisic 2002

Upload: thais

Post on 25-Feb-2016

36 views

Category:

Documents


0 download

DESCRIPTION

Introduction to Integer Programming Modeling and Methods. Michael Trick Carnegie Mellon University CPAI-OR School, Le Croisic 2002. Some History. Integer Programming goes back a long way: - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Introduction to Integer Programming Modeling and Methods

Introduction to Integer Programming Modeling and Methods

Michael TrickCarnegie Mellon UniversityCPAI-OR School, Le Croisic

2002

Page 2: Introduction to Integer Programming Modeling and Methods

Some History Integer Programming goes back a long way:

Schrijver takes it back to ancient times (linear diophantine equations), Euler (1748), Monge (1784) and much more.

“Proper” study began in the 1950s Dantzig (1951): linear programming Gomery (1958): cutting planes Land and Doig (1960): branch and bound

Survey books practically every five years since Tremendous practical success in last 10-15 years

Page 3: Introduction to Integer Programming Modeling and Methods

Scope This talk will not be comprehensive! Attempt to get across main

concepts of integer programming Relaxations Primal Heuristics Branch and Bound Cutting Planes

Page 4: Introduction to Integer Programming Modeling and Methods

Integer Program (IP)Minimize cxSubject to Ax=b l<=x<=u some or all of xj integral

X: variables

Linear objective

Linear constraints

Makes things hard!

Page 5: Introduction to Integer Programming Modeling and Methods

Rules of the Game Must put in that form! Seems limiting, but 50 years of

experience gives “tricks of the trade”

Many formulations for same problem

Page 6: Introduction to Integer Programming Modeling and Methods

Example FormulationsWarehouse location: n stores, m possible

warehouses; cost k[j] to open warehouse m; cost c[i,j] to handle store i out of warehouse j.

Minimize the total opening costs plus handling costs

Subject toEach store assigned to one open warehouse

Page 7: Introduction to Integer Programming Modeling and Methods

Warehouse FormulationVariables x[i,j] = 1 if store i served by warehouse

j; 0 otherwise y[j] = 1 if warehouse j open; 0

otherwise

Objective Minimize sum_j k[j]y[j]+sum_i,j c[i,j]x[i,j]

Page 8: Introduction to Integer Programming Modeling and Methods

Warehouse Formulation Constraints:

sum_j x[i,j] = 1 for all isum_i x[i,j] <= ny[j] for all j

Page 9: Introduction to Integer Programming Modeling and Methods

Binary Integer Programs Restrict variables to be 0-1 Many specialized methods OR people are real good at

formulating difficult problems within these restrictions

Page 10: Introduction to Integer Programming Modeling and Methods

Key concepts Relaxation R(IP)

“easily” solved problem such that Optimal solution value is no more than that of IP If solution is feasible for IP then is optimal for IP If R(IP) infeasible then so is IP

Most common is “linear relaxation”: drop integrality requirements and solve linear program

Others possible: lagrangian relaxation, lagrangian decomposition, bounds relaxation, etc.

Page 11: Introduction to Integer Programming Modeling and Methods

Illustration

Page 12: Introduction to Integer Programming Modeling and Methods

Linear Relaxation

Page 13: Introduction to Integer Programming Modeling and Methods

Why this fetish with Linear Relaxations? IP people are very focused on

linear relaxations. Why? Sometimes linear=integer Linear relaxations as global

constraints Duals and reduced costs

Page 14: Introduction to Integer Programming Modeling and Methods

Linear=integer formulations Happens naturally for some problems

Network flows Totally unimodular constraint matrices

Takes more work, but defined for Matchings Minimum spanning trees

Closely associated with polynomial solvability

Page 15: Introduction to Integer Programming Modeling and Methods

Duals and Reduced CostsAssociated with the solution of a

linear program are the dual values--- one per constraint--- measures the marginal value of changing the right-hand-side of the constraint

--- Useful in many algorithmic ways

Page 16: Introduction to Integer Programming Modeling and Methods

Dual exampleSum_i x[i,j]-y[j] <= 0

Suppose facility j* has cost 10 and y*[j*] = 0. The dual value of this constraint is 4. What

cost must facility j* have to be appealing?

Answer: no more than 10-4=6.

Page 17: Introduction to Integer Programming Modeling and Methods

Dual Example 2 Products 1, 2, 3 use chemicals A, B

Maximize 3x1+2x2+2x3Subject to x1+x2+2x3 <= 10; (.667) 5x1+2x2+x3 <= 20; (.667)

Solution: x2=6.67 x1=1.67What objective must a product that uses 4 of A

and 3 of B have to be appealing: at least 4.67

Page 18: Introduction to Integer Programming Modeling and Methods

Final advantage of linear relaxations: Global Linear relaxations are

Relatively easy to solve: huge advances in 15 years

Incorporate “global” information Often provide good bounds and guidelines

for integer program Variables with very bad reduced cost likely not in

optimal integer solution Rounding doesn’t always work, but often gets

good feasible solutions

Page 19: Introduction to Integer Programming Modeling and Methods

Feasible solutions Solutions that satisfy all the

constraints but might not be optimal Generally found by heuristics Can be problem specific

Must have value greater than or equal to optimal value (for minimizing)

Page 20: Introduction to Integer Programming Modeling and Methods

Feasible Solution

Feasible solution

Page 21: Introduction to Integer Programming Modeling and Methods

Fundamental Branch and Bound AlgorithmSolve relaxation to get x*If infeasible, then IP infeasibleElse If x* feasible to IP, then x*

optimal to IPElse create new problems IP1 and IP2

by branching; solve recursively, stop if prove subproblem cannot be optimal to IP (bounding)

Page 22: Introduction to Integer Programming Modeling and Methods

Branching Create two or more subproblems IP1, IP2,

… IPn such that Every feasible solution to IP appears in at

least one (often exactly one) of IP1, IP2, … IPn x* is infeasible to each of R(IP1), R(IP2), …

R(IPn) For linear relaxation, can choose a

fraction xj* and have one problem with xj <=[xj] and the other with xj >= [xj]+1 ([x]: round down of x)

Page 23: Introduction to Integer Programming Modeling and Methods

Illustration

x*

IP1

IP2

Page 24: Introduction to Integer Programming Modeling and Methods

Bounding Along way, we may find solution x’

that is feasible to IP. If any subproblem has relaxation value c* >= cx’ then we can prune that subproblem: it cannot contain the optimal solution. There is no sense continuing on that subproblem.

Page 25: Introduction to Integer Programming Modeling and Methods

Stopping Technique can stop early with

solution within a provable percentage of optimal (compare to be relaxation value)

Can also modify to generate all solutions (do not prune on ties)

Page 26: Introduction to Integer Programming Modeling and Methods

How to make work better? Better formulations Better relaxations (cuts) Better feasible solutions

(heuristics)

Page 27: Introduction to Integer Programming Modeling and Methods

Formulations Different formulations of integer

programs may act very differently: their relaxations might have radically different bounds

“Good Formulation” of integer program: provides a better relaxation value (all else being equal).

Page 28: Introduction to Integer Programming Modeling and Methods

Back to Warehouse Example Alternative formulation of “Only use if

open constraint”

x[i,j] <= y[j] for all i,j(versus)sum_i x[i,j] <= ny[j]

Which is better?

Page 29: Introduction to Integer Programming Modeling and Methods

Comparing Positives to original

Fewer constraints: linear relaxation should solve faster

Positives to disaggregate formulation Much better bounds (consider having x[i,j]=1

for a particular i,j. What would y[j] be in the two formulations?)

(Almost) no comparison! Formulation with more constraints works much better.

Page 30: Introduction to Integer Programming Modeling and Methods

Ideal

Formulation gives convex hull of feasible integer points

Page 31: Introduction to Integer Programming Modeling and Methods

Embarrassing Formulations Some things are very hard to formulate

with integer programming: Traveling Salesman problem: great success

story (IP approaches can optimize 15,000 city problems!), but best IP approaches begin with an exponentially sized formulation (no “good” compact formulation known).

Complicated operational requirements can be hard to formulate.

Page 32: Introduction to Integer Programming Modeling and Methods

Further approaches Branch and Price

Formulations with exponential number of variables with complexity in generating “good” variables (see Nemhauser): heavy use of dual values

Branch and Cut Improving formulations by adding

additional constraints to “cut off” linear relaxation solutions (more later)

Page 33: Introduction to Integer Programming Modeling and Methods

Algorithmic Details

PreprocessingPrimal HeuristicsBranchingCut Generation

Page 34: Introduction to Integer Programming Modeling and Methods

Preprocessing Process individual rows to

detect infeasibilities detect redundancies tighten right-hand-sides tighten variable bounds

Probing: examine consequences of fixing 0-1 variable If infeasible, fix to opposite bound If other variables are fixed, inequalities

Page 35: Introduction to Integer Programming Modeling and Methods

Preprocessing Much like simple Constraint

Programming

Page 36: Introduction to Integer Programming Modeling and Methods

Improving Coefficients3x1-2x2 1

1. Convert to with pos. coefficients with y1 = 1-x1

3y1 + 2x2 22. Note that constraint always satisfied when

y1 = 1, so change coefficient 3 to 22y1 + 2x2 2

3. Convert back to originalx1 -x2 1

Page 37: Introduction to Integer Programming Modeling and Methods

Improving (?) Coefficients

1

1

Cuts off (1/2, 1/4) and others

Page 38: Introduction to Integer Programming Modeling and Methods

Manual or Automatic? Modeling issue

Automatic identification not foolproof

Generally easy to see

Can provide problem-knowledge to further reduce coefficients

Automatic issue Many

opportunities will only occur within Branch and Bound tree as variables are fixed

“More foolproof” as models change

Page 39: Introduction to Integer Programming Modeling and Methods

Identifying Redundancy and InfeasibilityUse upper and lower bounds on variables: Redundancy

3x1 - 4x2 + 2x3 6 (max lhs is 5) Infeasibility

3x1 - 4x2 + 2x3 6 (max lhs is 5)

While very simple, can be used to fix variables, particularly within B&B tree

Page 40: Introduction to Integer Programming Modeling and Methods

PP: Fixing VariablesSimple idea: if setting a variable to a

value leads to infeasibility, then must set to another value

3x1-4x2+2x3-3x4 3

Setting x4 to 1 leads to previous infeasible constraint, so x4 must be 0

Page 41: Introduction to Integer Programming Modeling and Methods

PP: Implication Inequalities

Many constraints embed restrictions that at most one of x and y (or their complements) are 1.

This can lead to implication inequalities.

Page 42: Introduction to Integer Programming Modeling and Methods

PP: Implication InequalitiesFacility location

x1+x2+…+xm mx0

x0 = 0 x1 = 0x0 = 0 x2 = 0, etc.

(1-x0) + x1 1 (or x1 x0)x2 x0 , etc.

Automatic disaggregation (stronger!)

Page 43: Introduction to Integer Programming Modeling and Methods

PP: Clique InequalitiesThese inequalities found by “probing” (fix

variable and deduce implications).

These simple inequalities can be strengthened by combining mutually exclusive variables into a single constraint.

Resulting clique inequalities are very “strong” when many variables combined.

Page 44: Introduction to Integer Programming Modeling and Methods

Example: Sports SchedulingProblem: Given n teams, find an “optimal”

(minimum distance, equal distance, etc.) double round robin (every team plays at every other team) schedule.

A: @B @C D B C @DB: A D @C @A @D CC: @D A B D @A @BD: C @B @A @C B A

Page 45: Introduction to Integer Programming Modeling and Methods

Sports SchedulingMany formulations (not wholly

satisfactory)

One method: One variable for every “home stand” (series of home games) and “away trip” (series of away games).

Page 46: Introduction to Integer Programming Modeling and Methods

Variables (team A)Some variables:

2 3 41

H

@B @C

@C @D

H H

@E @F

y1x1

x2

y1

x3

Page 47: Introduction to Integer Programming Modeling and Methods

Constraints Can only do one thing in a time slot

y1+x1+x2 1x1+x2+y2 1

No “Away after Away”x1+x2+x3 1

No “Home after Home”y1+y2 1

Additional constraints link teams

Page 48: Introduction to Integer Programming Modeling and Methods

Improving FormulationCreate Implication Graph

2 3 41

H

@B @C

@C @D

H H

@E @F

y1x1

x2

y2

x3

Page 49: Introduction to Integer Programming Modeling and Methods

Find cliques Cliques in graph: can only have

one

2 3 41

H

@B @C

@C @D

H H

@E @F

y1x1

x2

y2

x3

Page 50: Introduction to Integer Programming Modeling and Methods

Constraints Can only do one thing in a time slot

y1+x1+x2 1x1+x2+y2 1

No “Away after Away”x1+x2+x3 + y2 1

No “Home after Home”y1+y2 + x1 + x2 1

Additional constraints link teams

Page 51: Introduction to Integer Programming Modeling and Methods

Clique Inequalities Resulting formulation is much tighter

(turns formulation from hopeless to possible)

Idea generalized to variables and their complements

Can be found automatically, but may be a huge number (and clique generally hard)

On divide of “automatic” and “manual” modeling issue

Page 52: Introduction to Integer Programming Modeling and Methods

Primal Heuristics Feasible solutions at B&B nodes

can greatly decrease the solution time

Most common: problem-specific heuristics embedded with B&B

Some general purpose heuristics: LP Diving, Pivot and Complement

Page 53: Introduction to Integer Programming Modeling and Methods

LP-Diving1. Solve LP2. Stop if infeasible or integral3. Fix all integral variables (or all 1’s)4. Select fractional variable and fix

to integer5. Go to 1

Page 54: Introduction to Integer Programming Modeling and Methods

Branching Two decisions: which B&B node to

branch on and how to divide into two problems

Node selection Depth First: try for integral solution Best Bound: explore “appealing” nodes Adaptive: Depth First first, then best

bound

Page 55: Introduction to Integer Programming Modeling and Methods

Branching: How to branch Priorities on variables and sets is

extremely important Normal branching is on a variable equals

0 or equals 1, but much more complicated branching possible:

x1 + x2 + x3 + x4 1Could lead to two problems:a) x1 + x2 = 0b) x3 + x4 =0

Page 56: Introduction to Integer Programming Modeling and Methods

Branching as ModelingSpecially Ordered Sets of Type 2 (no

more than 2 positive, must be adjacent): used to model piecewise linear functions:

x1,x2,x3,… xm

can lead to two problems:1 k-2 k-1 k+1 k+2 mk

Either all of the blue or all of the red are 0

Page 57: Introduction to Integer Programming Modeling and Methods

Adding constraints Constraints can strengthen

formulation: we have seen clique inequalities already

Can be added on an “as needed” basis during calculations (generally to move away from current fractional solution).

Page 58: Introduction to Integer Programming Modeling and Methods

Separation ProblemGiven a fractional solution to the LP

relaxation, find an inequality that is not satisfied by the fractional solution. Algorithm should be- fast- yield strong inequality

- heuristics acceptable

Page 59: Introduction to Integer Programming Modeling and Methods

Types of Constraints1. Feasibility: A large number of

constraints is used in formulation- connectivity constraints (i.e. TSP)

- nonlinearities2. Problem specific facets - blossom inequalities for matching - comb inequalities for TSP3. General IP constraints

Page 60: Introduction to Integer Programming Modeling and Methods

Gomory Constraints Any fractional solution from the simplex

method can be separated:xb + 2.5 x1 - 3.2 x2 = 2.1

where current sol. is xb =2.1, x1 = x2 = 0 xb + 2 x1 - 4 x2 - 2 = .1 - .5 x1 -.8 x2

LHS is integer so RHS must be also.1 - .5 x1 -.8 x2 0

is valid and violated by current solution

Page 61: Introduction to Integer Programming Modeling and Methods

“Real Version” of Gomory Cuts

y+ sum_j ajxj =d Let d=[d]+f; let aj=[aj] + fj t=y+sum_(j:fj<=f) [aj]xj + sum_(j:fj>f) ([aj]+1)xjSo d-t = sum_(j:fj<=f) fjxj+sum_(j:fj>f) (fj-1)xjEithert<=[d] => sum_(j:fj<=f) fjxj >= ft>=[d]+1 => sum_(j: fj>f) (1-fj)xj >= 1-fDivide through to get RHS of 1 in each case. Result givessum(j:fj<=f) (fj/f)xj + sum(j: fj>f) (1-fj)/(1-f)xj >= 1

Page 62: Introduction to Integer Programming Modeling and Methods

Working through examplePrevious example becomes

.5x1+.2x2 >= .9, which is a good, strong constraint

(can extend all this to mixed integer programs easily)

Page 63: Introduction to Integer Programming Modeling and Methods

Gomory Constraints (modeling) Are we done? Very good to have

available but likely not the only tool in the arsenal.

Page 64: Introduction to Integer Programming Modeling and Methods

0-1 Knapsack CoversFor problems that contain constraints like:

j N ajxj bC N is a cover if

j C aj > bthen

j C xj |C| - 1is valid

Page 65: Introduction to Integer Programming Modeling and Methods

Separation of Cover InequalitiesGiven fractional LP solution x*, is

there a cover C for which x* violates the cover inequality?

Solvable by a binary knapsack problem (one constraint IP): can be solved heuristically, or exactly by dynamic programming

Page 66: Introduction to Integer Programming Modeling and Methods

Lifting of Cover InequalitiesConstraints can be strengthened:

20x1+16x2+15x3+10x4+30x5 40Cover on {1,2,3} leads to

x1+x2+x3+0x4+0x5 2Is the “0” coefficient on x4 the best

possible? Can solve knapsack to getx1+x2+x3+x4+0x5 2

Page 67: Introduction to Integer Programming Modeling and Methods

Lifing of Cover InequalitiesWe could continue to x5 to get

x1+x2+x3+x4+x5 2If we had done x5 first, we would have

gotx1+x2+x3+0x4+2x5 2

Different lifting sequences lead to different inequalities

Page 68: Introduction to Integer Programming Modeling and Methods

Modeling implications Cover inequalities generally part of the

software, rather than the modeling (but if software has capability, do not include in model).

Decision is whether to use the inequalities or not essentially numerical question

Lifting is extremely important, as is relationship of covers in subproblems to full problem

Page 69: Introduction to Integer Programming Modeling and Methods

Conclusions There is lots to integer

programming! Interesting interplay between

formulations and algorithms Much intelligence embedded in

software, making formulations a bit more “fool-proof”

Page 70: Introduction to Integer Programming Modeling and Methods

Papers to ReadProgress in Linear Programming Based

Branch and Bound Algorithms: An exposition, Ellis Johnson, George Nemhauser, and Martin Savelsbergh

MIP: Theory and Practice – Closing the Gap, Robert Bixby, Mary Fenelon, Zonghao Gue, Ed Rothberg, and Roland Wunderling