On mathematical programming with indicator constraints
|
|
- Harry Bond
- 6 years ago
- Views:
Transcription
1 On mathematical programming with indicator constraints Andrea Lodi joint work with P. Bonami & A. Tramontani (IBM), S. Wiese (Unibo) University of Bologna, Italy École Polytechnique de Montréal, Québec, Canada International Symposium on Mathematical Programming (ISMP 2015) Pittsburgh, July 16, 2015 Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
2 Introduction & Motivation We consider Mathematical Optimization problems of the form min f (x, z) subject to h(x, z) 0 [z k = 0] [x S0 k] [z k = 1] [x S1 k] x R m z {0, 1} K. k = 1,..., K (P 0 ) The variable z k indicates whether or not the constraint sets S k 0 and Sk 1 are activated. Usually, these sets are defined as the intersection of level sets of given functions S = {x R m g i (x) 0, i = 1,..., l}. We deal with indicator constraints, allowed in Mathematical Programming software since the release of IBM-ILOG CPLEX 10. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
3 Introduction & Motivation (cont.d) Written as a Mathematical Programming problem (and slightly more general), (P 0 ) becomes min f (x, z) subject to h(x, z) 0 (x, z k ) Γ k j k = 1,..., K (P 1 ) j J x R m z J K, where Γ k j = S k j {j} and S k j R m j, k. This highlights the connection to the paradigm of Generalized Disjunctive Programming. In Mixed-Integer Programming there are (at least) two ways of dealing with the disjunctions associated with indicator constraints. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
4 The so-called bigm approach A very straightforward way to model an indicator constraint is g i (x) M i (1 z i ), where M i is a (very) large positive constant. By defining with F the feasible set for x, then M i sup x F g i (x). Although this quantity might not be easily (or not even at all) computable, setting a reasonably high value of M i will usually do the job in practice. However, two issues of this modeling trick are very well known: Weak continuous relaxations, and Severe numerical difficulties for both MILP & MINLP solvers. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
5 A change of perspective! In order to describe (x, z k ) Γ k j j J k = 1,..., K of (P 1 ), so as to be able of using the disjunctive programming approach for indicator constraints, we need the definition of the so-called perspective function. Definition 1 For a given closed convex function g R m R { }, the perspective function g R m+1 R { } is defined as λ g ( g(x, x λ) = λ ), λ > 0,, λ 0. (1) Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
6 Disjunctive convex optimization The perspective function is instrumental to extend the seminal work of Balas on the union of polyhedra to the study of the union of convex bodies. Namely, this extension has been studied by Ceria & Soares, whose main theorem is Theorem 1 (Ceria & Soares, 1999) Let C j = {x R m g j,i (x) 0, i = 1,..., l j }, for j J, and assume each g j,i R m R is a closed convex function. Then, conv ( j J C j ) = proj x cl(c), where (x, x,..., x J, λ 1,..., λ J ) R ( J +1)m+ J, x = j J x j, C = g j,i (x j, λ j ) 0, i = 1,..., l j, j J,. j J λ j = 1, λ j 0, j J Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
7 Disjunctive convex optimization (cont.d) The construction of Theorem 1 begins in a straightforward manner by creating copies x j of the initial variable x, each one constrained to lie in one of the disjunctive sets C j. Then, the original x is a convex combination of these copies with weights λ j. In this way, the resulting bilinear terms λ j x j lead to a set description of the convex hull that is, however, nonconvex. A simple transformation is then used to obtain the convex set C containing the perspective function. Although now the weakness of the continuous relaxation of the bigm approach seems solved by the strength of taking the convex hull in Theorem 1, Ceria & Soares result leaves a couple of highly nontrivial computational issues on the table. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
8 Disjunctive convex optimization: computational issues First, the closure of the perspective function has no algebraic representation in general. In other words, the perspective function becomes non-differentiable for λ j 0. The resulting numerical difficulties were already noticed in the process of designing a branch-and-cut algorithm for general 0 1 mixed convex programming problems, based on lift-and-project cuts separated through an alternative version of Theorem 1. [Stubbs & Mehrotra (1999)] It is worth noting that for more than ten years there was no significant progress in the separation of lift-and-project cuts for general mixed-integer convex programming problems. Recently, two major steps have been made to circumvent non-differentiability algorithmically through the solution of some (potentially many) easier optimization problems. [Bonami (2011) and Kılınç, Linderoth & Luedtke (2011)] Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
9 Issue #2: Lifting up and projecting down again The second difficulty associated with the practical use of Theorem 1 is the fact that the initial system is lifted into a space with a multiple dimension, i.e., C is defined in R ( J +1)m+ J, which makes the optimization over it computationally challenging. Although the representation of the convex hull is theoretically compact in the number of variables, on the practical side the size of the resulting NLPs had a significant impact on limiting the use of this disjunctive representation both in the separation of L&P cuts for convex MINLPs and, in general, for formulating MP problems with indicator constraints. However, in the context of indicator constraints, projecting out additional variables is possible in several special cases, as we will show in the following. (It is worth noting that working on the original space of variables had a huge impact also on L&P cuts in the linear case.) [Balas & Perregaard (2002) and Bonami (2012)] Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
10 Working on the original space of variables As mentioned working in the original space of variables is of fundamental importance in practice (and of mathematical fun). In the following we review some existing results in this area and we discuss new ones. Literature results are for the case of single disjunctions. Constraint vs nucleus On/off constraint On/off linear constraint The new results are instead for the case of pairs of related disjunctions. Complementary disjunctions Almost complementary disjunctions Ternary disjunctions Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
11 Single disjunctions: constraint vs. nucleus In the notation of (P 1 ) this special case can be written (dropping index k) as S 0 = {x R n x = 0}, and S 1 = {x R n l 1 x u 1, g 1,i (x) 0, i = 1,..., l}, and translates to Theorem 2 (Günlük & Linderoth, 2012) Let J = {0, 1} and for j J define Γ j = S j {j}, with S j specified as above and non-empty. We then have that if Γ 1 is a convex set, conv(γ 0 Γ 1 ) = cl(γ), where (x, z) R n+1, g 1,i (x, z) 0, i = 1,..., l, Γ =. zl 1 x zu 1, 0 < z 1 Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
12 Constraint vs. nucleus: example Example S 0 = {x R n x = 0}, and S 1 = {x R n l 1 x u 1, g 1,i (x) 0, i = 1,..., l}, let l 0 = l 1 = 0 R 2 and u 0 = u 1 = 1 R 2, and g 1 (x 1, x 2 ) = x 2 + 2(x ) 3 5. x 2 x 2 z x 1 x 1 Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
13 Single disjunctions: constraint vs. nucleus (cont.d) Note that in Theorem 2 the indicator variable z of the initial implication (or disjunction) takes over the role of the weight or multiplier inside the perspective function g i. Theorem 2 can also be formulated for the situation in which on the zero side, the variables are not forced to zero but to an arbitrary single point. Theorem 2 is also extendable to the situation in which on the zero side the set S 0 is not a single point, but a ray. Moreover, if the form of the functions g 1,i is known in more detail, in some cases one can avoid the need to take the closure and thus the differentiability issue, for example for polynomial functions. [Günlük & Linderoth (2012)] The remarkable computational advantages of Theorem 2 are, for example, for uncapacitated facility location with quadratic costs and unit commitment problems. [Günlük & Linderoth (2012) and Frangioni & Gentile (2006)] Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
14 Single disjunctions: on/off constraint Consider the slightly more general case where the set S 0 is not a point but a box. Namely, S 0 = {x R n l 0 x u 0 }, and S 1 = {x R n l 1 x u 1, g 1 (x) 0}. Theorem 3 (Hijazi et al., 2012) Let J = {0, 1} and for j J define Γ j = S j {j}, with S j specified as above and non-empty. We then have that if g 1 is isotone, conv(γ 0 Γ 1 ) = cl(γ ), where (x, z) R n+1, Γ zq 1,0 = I ( x, z) 0, I N, z zl 1 + (1 z)l 0 x zu 1 + (1 z)u 0, 0 < z 1. The statement of the theorem and the definition of Γ require the definition of an isotone function and its perspective q j,j I = g j h j,j I. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
15 A variant of the perspective for isotone functions Definition 2 A function g R n R is called independently non-decreasing (resp. non-increasing) in the i-th coordinate, if x dom(g) and λ > 0, we have g(x + λe i ) g(x) (resp. g(x + λe i ) g(x)). We say that g is independently monotone in the i-th coordinate, if it is either independently non-decreasing or non-increasing. Finally, g is called isotone, if it is independently monotone in every coordinate. For an isotone function g let J 1 (g) (resp. J 2 (g)) be the set of indices in which g is independently non-decreasing (resp. non-increasing), I N = {1,..., n}, x R n, z > 0 and j, j J (l j ) i i I J 1 (g j ), (u j ) i i I J 2 (g j ), I ) (x, z) = i x i (1 z)(u j ) i i z Ī J 1(g j ), x i (1 z)(l j ) i i z Ī J 2(g j ), (h j,j Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
16 On/off constraint: example Let l 0 = l 1 = 0 R 2 and u 0 = u 1 = 1 R 2, and consider the function g 0 (x 1, x 2 ) = exp (2x ) + x 2 1. x 2 x 2 z x 1 x 1 Formulating the delay-constrained routing problem by means of Theorem 3 leads to a significant computational advantage over a straightforward bigm formulation. [Hijazi et al. (2012)] Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
17 Theorem 3 vs Theorem 1: Trade-off On the one side, the advantage of Theorem 3 is the fact that we project back onto the original space of variables. In fact, we can express conv(γ 0 Γ 1 ) = cl(γ ) as a subset of a (n + 1)-dimensional space. A direct application of Theorem 1 would lead to expressing the convex hull as the projection of a subset of a (3n + 5)-dimensional space. On the other side, this gain does not come for free. In Theorem 3, we need exponentially many constraints. In particular, including all the simple bound constraints, we have 2 n + 2n + 2 constraints opposed to 4n + 9 that would result from a direct application of Theorem 1. In essence, an upper bound on the number of constraints needed to describe the convex hull is always exponential, although only in the number of variables involved in the constraint. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
18 Single disjunctions: on/off linear constraint In the special case that g 1 is an affine function there is no need of the closure. Corollary 1 (Hijazi, 2010) Consider the situation of Theorem 3. If g 1 is affine, conv(γ 0 Γ 1 ) = Γ, where (x, z) R n+1, Γ H 1,0 = I (x, z) 0, I N, zl 1 + (1 z)l 0 x zu 1 + (1 z)u 0, 0 z 1. where H j,j I (x, z) = i Ī a j,i x i + z a j,0 + (1 z) i I,a j,i >0 i Ī,a j,i >0 for any g j (x) = a j,0 + n i=1 a j,i x i, j J and I N. a j,i (l j ) i + a j,i (u j ) i + i I,a j,i <0 i Ī,a j,i <0 a j,i (u j ) i a j,i (l j ) i. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
19 On/off linear constraint: computation Supervised classification: consider a set Ω of m objects, where each object i Ω is characterized by the vector x i R d and associated with one of two classes, say, y i { 1, 1}. The task is to find a hyperplane ω T x b in R d that separates the two classes by maximizing a confidence margin. z i = 1 y i (ω T x i + b) 1 + ξ i 0 i = 1,..., m prep instance bigm lp CH lp P e bigm P e CH P e bigm P e CH weak... mean strong 2nl nl nl nl mean Trade-off: we have m vs m 2 d+2 constraints. However, many are redundant. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
20 Pairs of related disjunctions: complementary case We move on to two indicator constraints and therefore consider the sets S 0 = {x R n l 0 x u 0 }, S 1 = {x R n l 1 x u 1, g 1 (x) 0}, S 0 = {x R n l 0 x u 0, g 0 (x) 0}, S 1 = {x R n l 1 x u 1 }. We require that the two indicator constraints are complementary, i.e., the activation of one implies the deactivation of the other and vice versa. Because the two disjunctions above are complementary, i.e., z = 1 z, the whole situation could be described by a single disjunction with a single indicator variable, that is, as a subset of R n+1. For example, (x, z) ( S 0 {0}) (S 1 {1}). Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
21 Complementary disjunctions If the index sets of the two functions decompose in an opposite way, i.e., J 2 (g 1 ) = J 1 (g 0 ), then the above inequalities still give the convex hull. Theorem 4 (Bonami, Lodi, Tramontani & Wiese, 2015) Let Γ 0 = S 0 {0} and Γ 1 = S 1 {1}, with S 0 and S 1 specified as above and non-empty. Thus, if J 2 (g 1 ) = J 1 (g 0 ), then conv( Γ 0 Γ 1 ) = cl( Γ), where (x, z) R n+1, zq 1,0 I ( x, z) 0, z Γ = (1 z)q 0,1 I N, I ( x, 1 z) 0, 1 z zl 1 + (1 z)l 0 x zu 1 + (1 z)u 0, 0 < z < 1. In a sense this is a surprising result: conv ((Γ 0 Γ 1 ) ( Γ0 Γ 1 ) L = ) = conv (Γ 0 Γ 1 ) conv ( Γ0 Γ 1 ) L =, where L = = R n H = = {(z, z) z + z = 1}. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
22 Complementary disjunctions: example To the previous example add the function g 1 (x 1, x 2 ) = max { 1 2 x 1, 1 2 x 2}. x 2 x 2 z x 1 x 1 x 2 x 2 x 1 x 1 Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
23 Is J 2 (g 1 ) = J 1 (g 0 ) needed? Counterexample Let l 0 = l 1 = 0 and u 0 = u 1 = 1, H 0 (x) = x 1 2 and H1 (x) = x 3, both of which 4 are non-decreasing in their first and only coordinate. x x x z z z Open question: How to compute the convex hull in this case? Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
24 Complementary disjunctions: computation Job shop scheduling: we are given n jobs, J 1,..., J n, that have to be scheduled without preemption on m machines, M 1,..., M m. Because two jobs, say i, j, cannot overlap on a machine, say k, then x k ij = 1 s kj s ki + p ki x k ij = 0 s ki s kj + p kj (i, j, k) i < j instance bigm lp CH lp... P e bigm P e CH Pe bigm Pe CH la01-10x la03-10x la08-15x mean Trade-off: for each triple (i, j, k) we have 2 vs 8 constraints (reduced to 6). Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
25 Pairs of related disjunctions: almost complementary Instead of z + z = 1 as for complementary indicators, we could also consider z + z 1, leading to the three sets S 0 = {x R n l 0 x u 0 }, S 1 = {x R n l 1 x u 1, g 1 (x) 0}, and S 1 = {x R n l 1 x u 1, g 1 (x) 0}. In this case, we can compute a superset of the convex hull of the disjunction. We need a bit of notation: and for j J = {0, 1} and H = {(z, z) z + z 1} and L = R n H, Γ j = S j {j} [0, 1] and Γ j = S j [0, 1] {j}, (x, z, z) (Γ 0 Γ 1 ) ( Γ0 Γ 1 ) L. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
26 Almost complementary disjunctions With that we get the following Corollary: Corollary 2 (Bonami, Lodi, Tramontani & Wiese 2015) Let S 0, S 1 and S 1 be specified as above and non-empty. Thus, we have that if g 1 and g 1 are isotone functions, then conv ((Γ 0 Γ 1 ) ( Γ0 Γ 1 ) L ) conv (Γ 0 Γ 1 ) conv ( Γ0 Γ 1 ) L = cl( Γ), where Γ = (x, z, z) R n+2, zq 1,0 I ( x z, z) 0 zq 1,0 I ( x z, z) 0 0 < z, z < 1, z + z 1 I N, zl 1 + (1 z)l 0 x zu 1 + (1 z)u 0, zl 1 + (1 z)l 0 x zu 1 + (1 z)u 0,. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
27 Ternary disjunctions Almost complementary disjunctions could also be modeled by a single indicator variable assuming the values { 1, 0, 1}, i.e., ternary disjunctions. Natural questions are 1 is it possible to write the counterpart of Theorem 3 for this case? 2 which is the relationship between these two ways of representing essentially the same disjunction? Question 1 has a positive answer but only in the special case of affine functions because we are unable to define the perspective function for negative values of z. (The convexity-preserving property is otherwise lost.) The answer to Question 2 is also somehow surprising. We can show that, at least for affine functions, the convex hull of the almost complementary disjunction is contained in the convex hull of the ternary disjunction (when transforming everything into a common space). Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
28 Almost complementary disjunctions: computation Traveling salesman problem with time windows: find a minimum-cost tour visiting a set of cities exactly once, where each city i must be visited within a given time window [a i, b i ]. Binary variable x ij denotes if city i is visited in a tour immediately before city j, and activates the temporal constraint x ij = 1 s j s i + t ij (i, j) A instance bigmz lp CHz lp bigmx lp CHx lp P e bigmx P e CHx... rbg016a rbg040a rbg rbg201a *1.45 rbg *1.77 * mean Trade-off: for each pair (i, j) we have 2 versus 8 constraints (reduced to 6). Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
29 Computational facts, hopes and conclusions The computation is mildly promising although not conclusive because there is no instance that the general-purpose solver IBM-ILOG CPLEX can solve with the perspective formulation and cannot with the bigm one. However, by taking a closer look at the inequalities in the previous slides one realizes that, in the linear case, they have the structure of bigm constraints, and, in any case, they depend on the bounds on the involved variables. This opens the possibility of extensively using bound tightening in the nodes of the branch-and-bound tree where the perspective constraints can be replaced by locally-valid strengthened versions. This is what has been proposed to effectively solve supervised classification models and is now part of IBM-ILOG CPLEX algorithmic arsenal. [Belotti et al. (2014) and IBM-ILOG CPLEX (v )] We believe/hope that combining disjunctive formulations with aggressive cut generation and bound reductions within the tree could lead to effective algorithms for hard MINLPs. Andrea Lodi (Bologna U & Polytechnique Montréal) On MP with indicator constraints ISMP / 29
Indicator Constraints in Mixed-Integer Programming
Indicator Constraints in Mixed-Integer Programming Andrea Lodi University of Bologna, Italy - andrea.lodi@unibo.it Amaya Nogales-Gómez, Universidad de Sevilla, Spain Pietro Belotti, FICO, UK Matteo Fischetti,
More informationMixed Integer Programming Solvers: from Where to Where. Andrea Lodi University of Bologna, Italy
Mixed Integer Programming Solvers: from Where to Where Andrea Lodi University of Bologna, Italy andrea.lodi@unibo.it November 30, 2011 @ Explanatory Workshop on Locational Analysis, Sevilla A. Lodi, MIP
More informationInteger Programming ISE 418. Lecture 13. Dr. Ted Ralphs
Integer Programming ISE 418 Lecture 13 Dr. Ted Ralphs ISE 418 Lecture 13 1 Reading for This Lecture Nemhauser and Wolsey Sections II.1.1-II.1.3, II.1.6 Wolsey Chapter 8 CCZ Chapters 5 and 6 Valid Inequalities
More informationMixed Integer NonLinear Programs featuring On/Off constraints: convex analysis and applications
Mixed Integer NonLinear Programs featuring On/Off constraints: convex analysis and applications Hassan Hijai Pierre Bonami Gérard Cornuéjols Adam Ouorou October 2009 Abstract We call on/off constraint
More informationPERSPECTIVE REFORMULATION AND APPLICATIONS
PERSPECTIVE REFORMULATION AND APPLICATIONS OKTAY GÜNLÜK AND JEFF LINDEROTH Abstract. In this paper we survey recent work on the perspective reformulation approach that generates tight, tractable relaxations
More informationProjected Perspective Reformulations for NonLinear Network Design Problems
Projected Perspective Reformulations for NonLinear Network Design Problems Antonio Frangioni Claudio Gentile Enrico Grande Andrea Pacifici Dipartimento di Informatica, Università di Pisa Polo Universitario
More informationOn Handling Indicator Constraints in Mixed Integer Programming
On Handling Indicator Constraints in Mixed Integer Programming Pietro Belotti 1, Pierre Bonami 2, Matteo Fischetti 3, Andrea Lodi 4, Michele Monaci 3, Amaya Nogales-Gómez 5, and Domenico Salvagnin 3 1
More informationAdvances in CPLEX for Mixed Integer Nonlinear Optimization
Pierre Bonami and Andrea Tramontani IBM ILOG CPLEX ISMP 2015 - Pittsburgh - July 13 2015 Advances in CPLEX for Mixed Integer Nonlinear Optimization 1 2015 IBM Corporation CPLEX Optimization Studio 12.6.2
More informationConstraint Qualification Failure in Action
Constraint Qualification Failure in Action Hassan Hijazi a,, Leo Liberti b a The Australian National University, Data61-CSIRO, Canberra ACT 2601, Australia b CNRS, LIX, Ecole Polytechnique, 91128, Palaiseau,
More informationDisjunctive Cuts for Mixed Integer Nonlinear Programming Problems
Disjunctive Cuts for Mixed Integer Nonlinear Programming Problems Pierre Bonami, Jeff Linderoth, Andrea Lodi December 29, 2012 Abstract We survey recent progress in applying disjunctive programming theory
More informationOn handling indicator constraints in mixed integer programming
Comput Optim Appl (2016) 65:545 566 DOI 10.1007/s10589-016-9847-8 On handling indicator constraints in mixed integer programming Pietro Belotti 1 Pierre Bonami 2 Matteo Fischetti 3 Andrea Lodi 4,5 Michele
More informationMixed-Integer Nonlinear Programs featuring on/off constraints
Mixed-Integer Nonlinear Programs featuring on/off constraints Hassan Hijai Pierre Bonami Gérard Cornuéjols Adam Ouorou July 14, 2011 Abstract In this paper, we study MINLPs featuring on/off constraints.
More informationOn Counting Lattice Points and Chvátal-Gomory Cutting Planes
On Counting Lattice Points and Chvátal-Gomory Cutting Planes Andrea Lodi 1, Gilles Pesant 2, and Louis-Martin Rousseau 2 1 DEIS, Università di Bologna - andrea.lodi@unibo.it 2 CIRRELT, École Polytechnique
More informationDisjunctive Inequalities: Applications and Extensions
Disjunctive Inequalities: Applications and Extensions Pietro Belotti Leo Liberti Andrea Lodi Giacomo Nannicini Andrea Tramontani 1 Introduction A general optimization problem can be expressed in the form
More information15.081J/6.251J Introduction to Mathematical Programming. Lecture 24: Discrete Optimization
15.081J/6.251J Introduction to Mathematical Programming Lecture 24: Discrete Optimization 1 Outline Modeling with integer variables Slide 1 What is a good formulation? Theme: The Power of Formulations
More informationSolving Mixed-Integer Nonlinear Programs
Solving Mixed-Integer Nonlinear Programs (with SCIP) Ambros M. Gleixner Zuse Institute Berlin MATHEON Berlin Mathematical School 5th Porto Meeting on Mathematics for Industry, April 10 11, 2014, Porto
More informationLift-and-Project Inequalities
Lift-and-Project Inequalities Q. Louveaux Abstract The lift-and-project technique is a systematic way to generate valid inequalities for a mixed binary program. The technique is interesting both on the
More informationComputational Integer Programming. Lecture 2: Modeling and Formulation. Dr. Ted Ralphs
Computational Integer Programming Lecture 2: Modeling and Formulation Dr. Ted Ralphs Computational MILP Lecture 2 1 Reading for This Lecture N&W Sections I.1.1-I.1.6 Wolsey Chapter 1 CCZ Chapter 2 Computational
More informationProjected Chvátal-Gomory cuts for Mixed Integer Linear Programs. Pierre Bonami CMU, USA. Gerard Cornuéjols CMU, USA and LIF Marseille, France
Projected Chvátal-Gomory cuts for Mixed Integer Linear Programs Pierre Bonami CMU, USA Gerard Cornuéjols CMU, USA and LIF Marseille, France Sanjeeb Dash IBM T.J. Watson, USA Matteo Fischetti University
More informationInteger Programming ISE 418. Lecture 12. Dr. Ted Ralphs
Integer Programming ISE 418 Lecture 12 Dr. Ted Ralphs ISE 418 Lecture 12 1 Reading for This Lecture Nemhauser and Wolsey Sections II.2.1 Wolsey Chapter 9 ISE 418 Lecture 12 2 Generating Stronger Valid
More informationMixed-Integer Nonlinear Programming
Mixed-Integer Nonlinear Programming Claudia D Ambrosio CNRS researcher LIX, École Polytechnique, France pictures taken from slides by Leo Liberti MPRO PMA 2016-2017 Motivating Applications Nonlinear Knapsack
More informationAlternative Methods for Obtaining. Optimization Bounds. AFOSR Program Review, April Carnegie Mellon University. Grant FA
Alternative Methods for Obtaining Optimization Bounds J. N. Hooker Carnegie Mellon University AFOSR Program Review, April 2012 Grant FA9550-11-1-0180 Integrating OR and CP/AI Early support by AFOSR First
More informationScheduling and Optimization Course (MPRI)
MPRI Scheduling and optimization: lecture p. /6 Scheduling and Optimization Course (MPRI) Leo Liberti LIX, École Polytechnique, France MPRI Scheduling and optimization: lecture p. /6 Teachers Christoph
More informationA note on : A Superior Representation Method for Piecewise Linear Functions
A note on : A Superior Representation Method for Piecewise Linear Functions Juan Pablo Vielma Business Analytics and Mathematical Sciences Department, IBM T. J. Watson Research Center, Yorktown Heights,
More informationLMI Methods in Optimal and Robust Control
LMI Methods in Optimal and Robust Control Matthew M. Peet Arizona State University Lecture 02: Optimization (Convex and Otherwise) What is Optimization? An Optimization Problem has 3 parts. x F f(x) :
More informationSome Recent Advances in Mixed-Integer Nonlinear Programming
Some Recent Advances in Mixed-Integer Nonlinear Programming Andreas Wächter IBM T.J. Watson Research Center Yorktown Heights, New York andreasw@us.ibm.com SIAM Conference on Optimization 2008 Boston, MA
More informationPolyhedral Approach to Integer Linear Programming. Tepper School of Business Carnegie Mellon University, Pittsburgh
Polyhedral Approach to Integer Linear Programming Gérard Cornuéjols Tepper School of Business Carnegie Mellon University, Pittsburgh 1 / 30 Brief history First Algorithms Polynomial Algorithms Solving
More informationCutting planes from extended LP formulations
Cutting planes from extended LP formulations Merve Bodur University of Wisconsin-Madison mbodur@wisc.edu Sanjeeb Dash IBM Research sanjeebd@us.ibm.com March 7, 2016 Oktay Günlük IBM Research gunluk@us.ibm.com
More informationInteger Programming ISE 418. Lecture 8. Dr. Ted Ralphs
Integer Programming ISE 418 Lecture 8 Dr. Ted Ralphs ISE 418 Lecture 8 1 Reading for This Lecture Wolsey Chapter 2 Nemhauser and Wolsey Sections II.3.1, II.3.6, II.4.1, II.4.2, II.5.4 Duality for Mixed-Integer
More informationA note on : A Superior Representation Method for Piecewise Linear Functions by Li, Lu, Huang and Hu
A note on : A Superior Representation Method for Piecewise Linear Functions by Li, Lu, Huang and Hu Juan Pablo Vielma, Shabbir Ahmed and George Nemhauser H. Milton Stewart School of Industrial and Systems
More informationPseudo basic steps: Bound improvement guarantees from Lagrangian decomposition in convex disjunctive programming
Pseudo basic steps: Bound improvement guarantees from Lagrangian decomposition in convex disjunctive programming Dimitri J. Papageorgiou and Francisco Trespalacios Corporate Strategic Research ExxonMobil
More informationON MIXING SETS ARISING IN CHANCE-CONSTRAINED PROGRAMMING
ON MIXING SETS ARISING IN CHANCE-CONSTRAINED PROGRAMMING Abstract. The mixing set with a knapsack constraint arises in deterministic equivalent of chance-constrained programming problems with finite discrete
More informationStrong-Branching Inequalities for Convex Mixed Integer Nonlinear Programs
Computational Optimization and Applications manuscript No. (will be inserted by the editor) Strong-Branching Inequalities for Convex Mixed Integer Nonlinear Programs Mustafa Kılınç Jeff Linderoth James
More informationHow to Relax. CP 2008 Slide 1. John Hooker Carnegie Mellon University September 2008
How to Relax Slide 1 John Hooker Carnegie Mellon University September 2008 Two ways to relax Relax your mind and body. Relax your problem formulations. Slide 2 Relaxing a problem Feasible set of original
More informationEffective Separation of Disjunctive Cuts for Convex Mixed Integer Nonlinear Programs
Effective Separation of Disjunctive Cuts for Convex Mixed Integer Nonlinear Programs Mustafa Kılınç Jeff Linderoth James Luedtke Department of Industrial and Systems Engineering University of Wisconsin-Madison
More informationSolving Box-Constrained Nonconvex Quadratic Programs
Noname manuscript No. (will be inserted by the editor) Solving Box-Constrained Nonconvex Quadratic Programs Pierre Bonami Oktay Günlük Jeff Linderoth June 13, 2016 Abstract We present effective computational
More informationLift-and-Project Cuts for Mixed Integer Convex Programs
Lift-and-Project Cuts for Mixed Integer Convex Programs Pierre Bonami LIF, CNRS Aix-Marseille Université, 163 avenue de Luminy - Case 901 F-13288 Marseille Cedex 9 France pierre.bonami@lif.univ-mrs.fr
More informationChange in the State of the Art of (Mixed) Integer Programming
Change in the State of the Art of (Mixed) Integer Programming 1977 Vancouver Advanced Research Institute 24 papers 16 reports 1 paper computational, 4 small instances Report on computational aspects: only
More informationPseudo basic steps: Bound improvement guarantees from Lagrangian decomposition in convex disjunctive programming
Pseudo basic steps: Bound improvement guarantees from Lagrangian decomposition in convex disjunctive programming Dimitri J. Papageorgiou and Francisco Trespalacios Corporate Strategic Research ExxonMobil
More informationSeparation Techniques for Constrained Nonlinear 0 1 Programming
Separation Techniques for Constrained Nonlinear 0 1 Programming Christoph Buchheim Computer Science Department, University of Cologne and DEIS, University of Bologna MIP 2008, Columbia University, New
More informationFeasibility Pump for Mixed Integer Nonlinear Programs 1
Feasibility Pump for Mixed Integer Nonlinear Programs 1 Presenter: 1 by Pierre Bonami, Gerard Cornuejols, Andrea Lodi and Francois Margot Mixed Integer Linear or Nonlinear Programs (MILP/MINLP) Optimize
More informationExtended Formulations, Lagrangian Relaxation, & Column Generation: tackling large scale applications
Extended Formulations, Lagrangian Relaxation, & Column Generation: tackling large scale applications François Vanderbeck University of Bordeaux INRIA Bordeaux-Sud-Ouest part : Defining Extended Formulations
More informationComputational Experiments with Cross and Crooked Cross Cuts
Computational Experiments with Cross and Crooked Cross Cuts Sanjeeb Dash IBM Research sanjeebd@us.ibm.com Oktay Günlük IBM Research gunluk@us.ibm.com Juan Pablo Vielma Massachusetts Institute of Technology
More informationMixed Integer Linear Programming Formulations for Probabilistic Constraints
Mixed Integer Linear Programming Formulations for Probabilistic Constraints J. P. Vielma a,, S. Ahmed b, G. Nemhauser b a Department of Industrial Engineering, University of Pittsburgh 1048 Benedum Hall,
More informationChapter 3: Discrete Optimization Integer Programming
Chapter 3: Discrete Optimization Integer Programming Edoardo Amaldi DEIB Politecnico di Milano edoardo.amaldi@polimi.it Sito web: http://home.deib.polimi.it/amaldi/ott-13-14.shtml A.A. 2013-14 Edoardo
More informationTravelling Salesman Problem
Travelling Salesman Problem Fabio Furini November 10th, 2014 Travelling Salesman Problem 1 Outline 1 Traveling Salesman Problem Separation Travelling Salesman Problem 2 (Asymmetric) Traveling Salesman
More informationRelaxations of multilinear convex envelopes: dual is better than primal
of multilinear convex envelopes: dual is better than primal 1 LIX, École Polytechnique, Palaiseau, France June 7th, 2012 11 th International Symposium on Experimental Algorithms (SEA) 2012 - Bordeaux (France)
More informationApplications and algorithms for mixed integer nonlinear programming
Applications and algorithms for mixed integer nonlinear programming Sven Leyffer 1, Jeff Linderoth 2, James Luedtke 2, Andrew Miller 3, Todd Munson 1 1 Mathematics and Computer Science Division, Argonne
More informationThe Traveling Salesman Problem: An Overview. David P. Williamson, Cornell University Ebay Research January 21, 2014
The Traveling Salesman Problem: An Overview David P. Williamson, Cornell University Ebay Research January 21, 2014 (Cook 2012) A highly readable introduction Some terminology (imprecise) Problem Traditional
More informationSplit cuts and extended formulations for Mixed Integer Conic Quadratic Programming
Split cuts and extended formulations for Mixed Integer Conic Quadratic Programming The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.
More informationNumerical Optimization. Review: Unconstrained Optimization
Numerical Optimization Finding the best feasible solution Edward P. Gatzke Department of Chemical Engineering University of South Carolina Ed Gatzke (USC CHE ) Numerical Optimization ECHE 589, Spring 2011
More informationLecture 8: Column Generation
Lecture 8: Column Generation (3 units) Outline Cutting stock problem Classical IP formulation Set covering formulation Column generation A dual perspective Vehicle routing problem 1 / 33 Cutting stock
More informationLift-and-project cuts for convex mixed integer nonlinear programs. Linear programming based separation and extended formulations
Math. Prog. Comp. DOI 10.1007/s12532-017-0118-1 FULL LENGTH PAPER 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Lift-and-project cuts for convex mixed integer nonlinear programs Linear programming based separation
More informationOptimizing Convex Functions over Non-Convex Domains
Optimizing Convex Functions over Non-Convex Domains Daniel Bienstock and Alexander Michalka University Berlin 2012 Introduction Generic Problem: min Q(x), s.t. x F, Introduction Generic Problem: min Q(x),
More informationAn Exact Algorithm for the Steiner Tree Problem with Delays
Electronic Notes in Discrete Mathematics 36 (2010) 223 230 www.elsevier.com/locate/endm An Exact Algorithm for the Steiner Tree Problem with Delays Valeria Leggieri 1 Dipartimento di Matematica, Università
More informationOn the Relative Strength of Split, Triangle and Quadrilateral Cuts
On the Relative Strength of Split, Triangle and Quadrilateral Cuts Amitabh Basu Tepper School of Business, Carnegie Mellon University, Pittsburgh, PA 53 abasu@andrew.cmu.edu Pierre Bonami LIF, Faculté
More informationSymmetry in Scheduling Problems
Symmetry in Scheduling Problems James Ostrowski 1, Miguel F. Anjos 2, and Anthony Vannelli 3 1 Decision and Information Sciences, Argonne National Lab, Argonne, IL, jostrowski@anl.gov 2 GERAD & Département
More informationIntroduction to Mathematical Programming IE406. Lecture 21. Dr. Ted Ralphs
Introduction to Mathematical Programming IE406 Lecture 21 Dr. Ted Ralphs IE406 Lecture 21 1 Reading for This Lecture Bertsimas Sections 10.2, 10.3, 11.1, 11.2 IE406 Lecture 21 2 Branch and Bound Branch
More informationLifted Inequalities for 0 1 Mixed-Integer Bilinear Covering Sets
1 2 3 Lifted Inequalities for 0 1 Mixed-Integer Bilinear Covering Sets Kwanghun Chung 1, Jean-Philippe P. Richard 2, Mohit Tawarmalani 3 March 1, 2011 4 5 6 7 8 9 Abstract In this paper, we study 0 1 mixed-integer
More informationOn the knapsack closure of 0-1 Integer Linear Programs
On the knapsack closure of 0-1 Integer Linear Programs Matteo Fischetti 1 Dipartimento di Ingegneria dell Informazione University of Padova Padova, Italy Andrea Lodi 2 Dipartimento di Elettronica, Informatica
More informationInteger Programming ISE 418. Lecture 13b. Dr. Ted Ralphs
Integer Programming ISE 418 Lecture 13b Dr. Ted Ralphs ISE 418 Lecture 13b 1 Reading for This Lecture Nemhauser and Wolsey Sections II.1.1-II.1.3, II.1.6 Wolsey Chapter 8 CCZ Chapters 5 and 6 Valid Inequalities
More informationChapter 3: Discrete Optimization Integer Programming
Chapter 3: Discrete Optimization Integer Programming Edoardo Amaldi DEIB Politecnico di Milano edoardo.amaldi@polimi.it Website: http://home.deib.polimi.it/amaldi/opt-16-17.shtml Academic year 2016-17
More informationOn the Relative Strength of Split, Triangle and Quadrilateral Cuts
On the Relative Strength of Split, Triangle and Quadrilateral Cuts Amitabh Basu Pierre Bonami Gérard Cornuéjols François Margot Abstract Integer programs defined by two equations with two free integer
More informationThe Chvátal-Gomory Closure of an Ellipsoid is a Polyhedron
The Chvátal-Gomory Closure of an Ellipsoid is a Polyhedron Santanu S. Dey 1 and Juan Pablo Vielma 2,3 1 H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology,
More informationFundamental Domains for Integer Programs with Symmetries
Fundamental Domains for Integer Programs with Symmetries Eric J. Friedman Cornell University, Ithaca, NY 14850, ejf27@cornell.edu, WWW home page: http://www.people.cornell.edu/pages/ejf27/ Abstract. We
More informationOnline generation via offline selection - Low dimensional linear cuts from QP SDP relaxation -
Online generation via offline selection - Low dimensional linear cuts from QP SDP relaxation - Radu Baltean-Lugojan Ruth Misener Computational Optimisation Group Department of Computing Pierre Bonami Andrea
More informationDisjunctive Cuts for Cross-Sections of the Second-Order Cone
Disjunctive Cuts for Cross-Sections of the Second-Order Cone Sercan Yıldız Gérard Cornuéjols June 10, 2014 Abstract In this paper we provide a unified treatment of general two-term disjunctions on crosssections
More informationTwo-Term Disjunctions on the Second-Order Cone
Noname manuscript No. (will be inserted by the editor) Two-Term Disjunctions on the Second-Order Cone Fatma Kılınç-Karzan Sercan Yıldız the date of receipt and acceptance should be inserted later Abstract
More informationOutline. 1 Introduction. 2 Modeling and Applications. 3 Algorithms Convex. 4 Algorithms Nonconvex. 5 The Future of MINLP? 2 / 126
Outline 1 Introduction 2 3 Algorithms Convex 4 Algorithms Nonconvex 5 The Future of MINLP? 2 / 126 Larry Leading Off 3 / 126 Introduction Outline 1 Introduction Software Tools and Online Resources Basic
More informationMixed Integer Non Linear Programming
Mixed Integer Non Linear Programming Claudia D Ambrosio CNRS Research Scientist CNRS & LIX, École Polytechnique MPRO PMA 2016-2017 Outline What is a MINLP? Dealing with nonconvexities Global Optimization
More informationLogic, Optimization and Data Analytics
Logic, Optimization and Data Analytics John Hooker Carnegie Mellon University United Technologies Research Center, Cork, Ireland August 2015 Thesis Logic and optimization have an underlying unity. Ideas
More informationOperations Research Methods in Constraint Programming
Operations Research Methods in Constraint Programming John Hooker Carnegie Mellon University Prepared for Lloret de Mar, Spain June 2007 2007 Slide 1 CP and OR Have Complementary Strengths CP: Inference
More informationOptimization in Process Systems Engineering
Optimization in Process Systems Engineering M.Sc. Jan Kronqvist Process Design & Systems Engineering Laboratory Faculty of Science and Engineering Åbo Akademi University Most optimization problems in production
More informationSection #2: Linear and Integer Programming
Section #2: Linear and Integer Programming Prof. Dr. Sven Seuken 8.3.2012 (with most slides borrowed from David Parkes) Housekeeping Game Theory homework submitted? HW-00 and HW-01 returned Feedback on
More informationMVE165/MMG631 Linear and integer optimization with applications Lecture 8 Discrete optimization: theory and algorithms
MVE165/MMG631 Linear and integer optimization with applications Lecture 8 Discrete optimization: theory and algorithms Ann-Brith Strömberg 2017 04 07 Lecture 8 Linear and integer optimization with applications
More informationA new family of facet defining inequalities for the maximum edge-weighted clique problem
A new family of facet defining inequalities for the maximum edge-weighted clique problem Franklin Djeumou Fomeni June 2016 Abstract This paper considers a family of cutting planes, recently developed for
More informationSplit Rank of Triangle and Quadrilateral Inequalities
Split Rank of Triangle and Quadrilateral Inequalities Santanu Dey 1 Quentin Louveaux 2 June 4, 2009 Abstract A simple relaxation of two rows of a simplex tableau is a mixed integer set consisting of two
More informationThe Split Closure of a Strictly Convex Body
The Split Closure of a Strictly Convex Body D. Dadush a, S. S. Dey a, J. P. Vielma b,c, a H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive
More informationDecomposition-based Methods for Large-scale Discrete Optimization p.1
Decomposition-based Methods for Large-scale Discrete Optimization Matthew V Galati Ted K Ralphs Department of Industrial and Systems Engineering Lehigh University, Bethlehem, PA, USA Départment de Mathématiques
More informationCutting Planes and Elementary Closures for Non-linear Integer Programming
Cutting Planes and Elementary Closures for Non-linear Integer Programming Juan Pablo Vielma University of Pittsburgh joint work with D. Dadush and S. S. Dey S. Modaresi and M. Kılınç Georgia Institute
More informationMVE165/MMG630, Applied Optimization Lecture 6 Integer linear programming: models and applications; complexity. Ann-Brith Strömberg
MVE165/MMG630, Integer linear programming: models and applications; complexity Ann-Brith Strömberg 2011 04 01 Modelling with integer variables (Ch. 13.1) Variables Linear programming (LP) uses continuous
More informationJeff Howbert Introduction to Machine Learning Winter
Classification / Regression Support Vector Machines Jeff Howbert Introduction to Machine Learning Winter 2012 1 Topics SVM classifiers for linearly separable classes SVM classifiers for non-linearly separable
More informationA Branch-and-Refine Method for Nonconvex Mixed-Integer Optimization
A Branch-and-Refine Method for Nonconvex Mixed-Integer Optimization Sven Leyffer 2 Annick Sartenaer 1 Emilie Wanufelle 1 1 University of Namur, Belgium 2 Argonne National Laboratory, USA IMA Workshop,
More informationPART 4 INTEGER PROGRAMMING
PART 4 INTEGER PROGRAMMING 102 Read Chapters 11 and 12 in textbook 103 A capital budgeting problem We want to invest $19 000 Four investment opportunities which cannot be split (take it or leave it) 1.
More informationLecture 4: NP and computational intractability
Chapter 4 Lecture 4: NP and computational intractability Listen to: Find the longest path, Daniel Barret What do we do today: polynomial time reduction NP, co-np and NP complete problems some examples
More informationMonoidal Cut Strengthening and Generalized Mixed-Integer Rounding for Disjunctions and Complementarity Constraints
Monoidal Cut Strengthening and Generalized Mixed-Integer Rounding for Disjunctions and Complementarity Constraints Tobias Fischer and Marc E. Pfetsch Department of Mathematics, TU Darmstadt, Germany {tfischer,pfetsch}@opt.tu-darmstadt.de
More informationComputational Experiments with Cross and Crooked Cross Cuts
Submitted to INFORMS Journal on Computing manuscript (Please, provide the mansucript number!) Authors are encouraged to submit new papers to INFORMS journals by means of a style file template, which includes
More informationA notion of Total Dual Integrality for Convex, Semidefinite and Extended Formulations
A notion of for Convex, Semidefinite and Extended Formulations Marcel de Carli Silva Levent Tunçel April 26, 2018 A vector in R n is integral if each of its components is an integer, A vector in R n is
More informationEE/ACM Applications of Convex Optimization in Signal Processing and Communications Lecture 18
EE/ACM 150 - Applications of Convex Optimization in Signal Processing and Communications Lecture 18 Andre Tkacenko Signal Processing Research Group Jet Propulsion Laboratory May 31, 2012 Andre Tkacenko
More informationOptimizing the Efficiency of the Liver Allocation System through Region Selection
Optimizing the Efficiency of the Liver Allocation System through Region Selection Nan Kong Department of Industrial and Management Systems Engineering, University of South Florida Andrew J. Schaefer, Brady
More informationA Time Bucket Formulation for the TSP with Time Windows
A Time Bucket Formulation for the TSP with Time Windows Sanjeeb Dash, Oktay Günlük IBM Research Andrea Lodi, Andrea Tramontani University of Bologna November 10, 2009 Abstract The Traveling Salesman Problem
More informationA disjunctive convex programming approach to the pollution routing problem
A disjunctive convex programming approach to the pollution routing problem Ricardo Fukasawa, Qie He, Yongjia Song Abstract The pollution routing problem (PRP) aims to determine a set of routes and speed
More informationMinimizing Cubic and Homogeneous Polynomials over Integers in the Plane
Minimizing Cubic and Homogeneous Polynomials over Integers in the Plane Alberto Del Pia Department of Industrial and Systems Engineering & Wisconsin Institutes for Discovery, University of Wisconsin-Madison
More informationBCOL RESEARCH REPORT 07.04
BCOL RESEARCH REPORT 07.04 Industrial Engineering & Operations Research University of California, Berkeley, CA 94720-1777 LIFTING FOR CONIC MIXED-INTEGER PROGRAMMING ALPER ATAMTÜRK AND VISHNU NARAYANAN
More informationIntersection cuts for nonlinear integer programming: convexification techniques for structured sets
Noname manuscript No. (will be inserted by the editor Intersection cuts for nonlinear integer programming: convexification techniques for structured sets Sina Modaresi Mustafa R. Kılınç Juan Pablo Vielma
More informationModified Orbital Branching with Applications to Orbitopes and to Unit Commitment
Modified Orbital Branching with Applications to Orbitopes and to Unit Commitment James Ostrowski Miguel F. Anjos Anthony Vannelli October 27, 2012 Abstract The past decade has seen advances in general
More informationMIP reformulations of some chance-constrained mathematical programs
MIP reformulations of some chance-constrained mathematical programs Ricardo Fukasawa Department of Combinatorics & Optimization University of Waterloo December 4th, 2012 FIELDS Industrial Optimization
More informationCS6999 Probabilistic Methods in Integer Programming Randomized Rounding Andrew D. Smith April 2003
CS6999 Probabilistic Methods in Integer Programming Randomized Rounding April 2003 Overview 2 Background Randomized Rounding Handling Feasibility Derandomization Advanced Techniques Integer Programming
More informationThe Split Closure of a Strictly Convex Body
The Split Closure of a Strictly Convex Body D. Dadush a, S. S. Dey a, J. P. Vielma b,c, a H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive
More informationOn mixed-integer sets with two integer variables
On mixed-integer sets with two integer variables Sanjeeb Dash IBM Research sanjeebd@us.ibm.com Santanu S. Dey Georgia Inst. Tech. santanu.dey@isye.gatech.edu September 8, 2010 Oktay Günlük IBM Research
More information