Facets from Gadgets. Adam N. Letchford Anh Vu. December 2017

Size: px
Start display at page:

Download "Facets from Gadgets. Adam N. Letchford Anh Vu. December 2017"

Transcription

1 Facets from Gadgets Adam N. Letchford Anh Vu December 2017 Abstract We present a new tool for generating cutting planes for N P-hard combinatorial optimisation problems. It is based on the concept of gadgets small subproblems that are glued together to form hard problems which we borrow from the literature on computational complexity. Using gadgets, we are able to derive huge (exponentially large) new families of strong (and sometimes facet-defining) cutting planes, accompanied by efficient separation algorithms. We illustrate the power of this approach on the stable set and clique partitioning problems. Keywords: branch-and-cut, gadgets, stable set problem, clique partitioning problem. 1 Introduction A popular approach to solving N P-hard combinatorial optimisation problems exactly is branch-and-cut [30], in which valid linear inequalities (also called cutting planes or simply cuts) are used to strengthen linear programming relaxations of the problem. Algorithms for generating cuts are called separation algorithms, and the strongest possible cuts are ones that define facets of the associated polyhedron (see, e.g., [9, 22]). The purpose of this paper is to present a new tool for generating cutting planes. It is based on the concept of gadgets small subproblems that are glued together to form hard problems which we borrow from the literature on computational complexity theory (see, e.g., [13, 33]). Using gadgets, we are able to derive huge (exponentially large) new families of strong cutting planes. We also obtain efficient separation algorithms as a by-product of the analysis. We illustrate the power of the gadget approach on two well-known and much-studied N P-hard problems: the stable set problem [16] and the clique Corresponding author. Department of Management Science, Lancaster University, Lancaster LA1 4YX, United Kingdom. a.n.letchford@lancaster.ac.uk Also at Lancaster. a.vu@lancaster.ac.uk 1

2 partitioning problem [18, 25]. Interestingly, for both problems, some of the new cutting planes that we derive are not only facet-defining, but also have rank greater than one, in the sense of Chvátal [8]. The structure of the paper is as follows. In Section 2, we review the relevant literature. In Section 3, we define gadgets and show how they can be used both to derive cuts and solve the associated separation problem. The applications to the stable set and clique partitioning problems are in Sections 4 and 5, respectively. Finally, in Section 6, we make some remarks about the possibility of applying our approach to other combinatorial optimisation problems. Throughout the paper, we use the following (standard) graph-theoretic notation and terminology. G = (V, E) will denote a generic (simple, loopless) undirected graph. The complement of G is the graph with node set V and edge set { {i, j} V : {i, j} / E }. Given a node i V, we let n(i) denote the neighbours of i in G, i.e., the set of nodes adjacent to i. Given a set S V, we let G[S] denote the subgraph of G induced by S. A clique (respectively, stable set) is a set of pairwise adjacent (respectively, non-adjacent) nodes. A cycle is a connected subgraph in which all nodes have degree two. Given a cycle C, we let V (C) denote its vertex set and E(C) its edge set. Given a positive integer n, we let K n = (V n, E n ) denote the complete undirected graph on n nodes, where V n = {1,..., n} and E n = {e V n : e = 2}. A directed graph, cycle or path will be called a digraph, dicycle or dipath, respectively. In a digraph, nodes are linked by arcs rather than edges. 2 Literature Review We now review the relevant literature. For reasons which will become apparent, we start by describing the satisfiability problem in Subsection 2.1. We then cover gadgets in Subsection 2.2. The remaining two subsections concern the stable set and clique partitioning problems. 2.1 Satisfiability The satisfiability problem (SAT) was the first ever problem to be proved N P- complete [10]. In this problem, we have a collection of Boolean variables, say x 1,..., x n, and a collection of logical disjunctions, called clauses, that they should satisfy. An example of a clause is x 1 x 2 x 3, which should be read as x 1 is true or x 2 is false or x 3 is false. In this context, or is always intended to be inclusive. The task is to check whether there exists an assignment of truth values to the variables that satisfies all of the clauses. The special case of SAT in which every clause contains exactly k terms is called k-sat. It is well known that 2-SAT can be solved in linear time. One way to do it as as follows [1]. Replace each clause with two implications. For example, the clause x 1 x 2 is replaced with the statements x 1 implies 2

3 x 2 and x 2 implies x 1. Construct a digraph with 2n nodes, where the first n nodes represent x 1,..., x n and the others represent x 1,..., x n. Represent every implication by a directed arc. For example, the first of the above implications is represented by an arc from node x 1 to node x 2. Compute the strongly connected components of the digraph. The instance is satisfiable if and only there does not exist an index i such that both node x i and node x i lie in the same component. 2.2 Gadgets Gadgets (also sometimes called units or components) are used heavily in computational complexity theory, in order prove hardness results. For example, suppose we wish to prove that 3-SAT is N P-complete. Since we know that SAT is N P-complete, it suffices to show that any clause with more than 3 terms is equivalent to a small (polynomially bounded) number of clauses with only 3 terms (possibly with the help of additional variables). This is indeed the case. For example, suppose we have a clause of the form p i=1 x i, with p 4. Let q = p/2 and replace the clause with two shorter clauses of the form y q i=1 x i and ȳ p i=q+1 x i, where y is an additional Boolean variable. Using this gadget repeatedly, if necessary, we eventually obtain O(k) clauses with only three terms each, using only O(k) additional variables. Instead of giving more details, we refer the reader to [13, 33]. 2.3 The stable set problem Let G be an undirected graph. The maximum cardinality of a stable set in G is called the stability number (or independence number) of G and denoted by α(g). The stable set problem (SSP) is the problem of computing α(g). A generalisation is the maximum weight stable set problem, in which are also given a weight vector w Q n +, and seek a stable set of maximum total weight. The SSP is equivalent to the maximum clique problem on the complement of G. The SSP is therefore not only N P-hard in the strong sense, but hard even to approximate [21]. The simplest formulation of the SSP is the following. For each i v, let x i be a binary variable, taking the value 1 if and only if node i is selected. Then: max i V x i (1) s.t. x i + x j 1 ({i, j} E) (2) x i {0, 1} (i V ). (3) The constraints (2) are called edge inequalities. The convex hull of feasible solutions to (1) (3) is called the stable set polytope and denoted by STAB(G). A wide range of valid inequalities are 3

4 3 2 Figure 1: Odd hole inequality (left) and odd antihole inequality (right). known for this polytope (see the survey [16]). For example, Padberg [29] introduced the following inequalities: clique inequalities i C x i 1 for each maximal clique C V ; odd hole inequalities i S x i S /2 for each set of nodes S inducing an odd hole (chordless cycle of odd length) in G; odd antihole inequalities i S x i 2 for each set of nodes S inducing the complement of an odd hole in G. See Fig. 1 for a representation of odd hole and odd antihole inequalities with S = 7. Small circles represent nodes in S and lines represent edges in G[S]. Padberg showed that the clique inequalities define facets of STAB(G). He also showed that any odd hole or odd antihole inequality either defines a facet, or can be lifted (strengthened) to obtain a facet-defining inequality. The separation problem for the clique inequalities is N P-hard, but some effective separation heuristics exist [27, 32]. The separation problem for the odd hole inequalities can be solved in polynomial time, by reduction to a series of V shortest (s, t)-path problems in a graph that is twice the size of G [14]. If we use the Fibonacci heap variant of Dijkstra s algorithm [12] to solve those ( shortest-path problems, ) the total time taken for odd hole separation is O V E + V 2 log V. Separation algorithms for other inequalities can be found in, e.g., [2, 4, 5, 15, 16, 24, 32]. We omit details for brevity. 2.4 The clique partitioning problem The clique partitioning problem (CPP) was first defined by Marcotorchino [25]. We are given an integer n 2 and a rational weight w e for each edge e E n. The task is to partition V n into subsets (cliques), such that the sum of the weights of the edges that have both end-nodes in the same clique is maximised. The CPP was proved to be N P-hard in the strong sense in [34]. 4

5 3 2 Figure 2: A 2-chorded odd cycle inequality (left) and an odd wheel inequality (right). Marcotorchino formulated the CPP as the following 0-1 LP: max e E n w e x e (4) s.t. x ik + x jk x ij 1 ({i, j} E n, k V n \ {i, j}) (5) x e {0, 1} (e E n ). (6) Here, the binary variable x ij takes the value 1 if and only if nodes i and j are in the same clique. The inequalities (5) are called transitivity inequalities. The convex hull of feasible solutions to (4) (6) is called the clique partitioning polytope and denoted by CPP n. Grötschel and Wakabayashi [19] showed that the transitivity inequalities (5), and the non-negativity inequalities x e 0 for all e E n, define facets of CPP n. They also introduced several other families of facet-defining inequalities. Of those, we will be especially interested in the 2-chorded odd cycle (2-COC) inequalities. These take the form: x e C /2, (7) e E(C) x e e H where C is a cycle in K n with C 5 and odd, and H is the set of 2-chords of C, i.e., the set of edges in E n which connect pairs of nodes in V (C) that have a distance 2 in C. Chopra and Rao [6] found two additional families of facet-defining inequalities for CPP n. Of those, we will be interested in the odd wheel (OW) inequalities. These take the form: x vh x e C /2, (8) v V (C) e E(C) where C is a cycle in K n with C 3 and odd, and h is any arbitrary node in V n \ V (C). Fig. 2 gives a representation of a 2-COC inequality with C = 7 and an OW inequality with C = 5. Solid and dotted lines represent edges whose variables have coefficient 1 and 1, respectively. 5

6 e(7) f(6) f(7) e(1) e(2) f(1) f(5) f(4) e(6) e(4) f(2) e(3) f(3) 3 e(5) Figure 3: An OCW inequality with c = 7. Müller [26] and Caprara & Fischetti [3] independently discovered a family of valid inequalities that includes the 2-COC and OW inequalities as special cases. We will follow Müller in calling them odd closed walk (OCW) inequalities. Suppose we are given a sequence e(1),..., e(c) of edges in E n such that c 3 and odd and such that, for i = 1,..., c, edge e(i) and edge e(i + 1) have exactly one end-node in common (indices taken modulo c). Also, for i = 1,..., c, let f(i) be the unique edge that shares one end-node with e(i) and the other end-node with e(i + 1). The OCW inequality takes the form: c c x e(i) x f(i) c/2. (9) i=1 i=1 Fig. 3 gives a representation of an OCW inequality that is not a 2-COC or OW inequality. Deza et al. [11] showed that the separation problem for the OW inequalities can be solved in O ( n 4) time, by reduction to n 2 shortest-path computations in K n 1. The complexity of separation for the 2-COC inequalities is unknown, but there exists an efficient separation algorithm for the more general OCW inequalities [3, 26]. The algorithm involves the solution of O ( n 2) shortest-path problems in a graph with O ( n 2) nodes and O ( n 3) edges. Using the algorithm in [12] to solve these shortest-path problems, one obtains a total running time of O ( n 5). Separation algorithms for other inequalities can be found in, e.g., [2, 11, 18, 23, 28]. Again, we omit details, for brevity. 3 Using Gadgets to Generate Cutting Planes In this section, we present our gadget approach in its general form. In Subsection 3.1, we define our gadgets formally. In Subsection 3.2, we show how they can be used to generate cuts. Subsections 3.3 and 3.4 are devoted to exact separation and heuristic separation, respectively. Finally, in Subsec- 6

7 tion 3.5, we relate our approach to two other general schemes for generating cuts. 3.1 Gadgets Consider any combinatorial optimisation problem that can be formulated as an integer linear program (ILP). Let S Z n + be the set of feasible solutions and let P = conv(s) be the associated polytope. We will need the following two definitions. Definition 1 Let α T x β be any valid linear inequality that defines a proper face of P. A point x S satisfying α T x = β will be called a root of the inequality. Definition 2 Let Z 0 and Z 1 denote the non-negative even and odd integers, respectively. A valid inequality α T x β will be called a gadget if it satisfies the following properties: it has integral coefficients; it defines a proper face of P ; there exists a pair {i, j} {1,..., n} and a pair p, q {0, 1} such that all roots of the inequality with x i Z p also have x j Z q. We will say that the pair {i, j} is associated with the given gadget. The following example shows that it is possible for several different pairs of variables to be associated with the same gadget. Example 1 Suppose the inequality x 1 + x 2 + 3x 3 3 is valid for P. Then all roots of the inequality with x 1 odd have x 2 and x 3 even, all roots with x 2 odd have x 1 and x 3 even, and all roots with x 3 odd have both x 1 and 2 even. Thus, the given inequality is a gadget, and it is associated with the pairs {1, 2}, {1, 3} and {2, 3}. Two special kinds of gadget will prove to be useful later on: Definition 3 A gadget will be called an exclusive or (XOR) gadget if there exists a pair {i, j} {1,..., n} such that, in all roots, exactly one of the variables x i and x j is odd. Definition 4 A gadget will be called an equality (EQ) gadget if there exists a pair {i, j} {1,..., n} such that, in all roots, the variables x i and x j have the same parity. Example 2 Suppose the inequality x 1 +2x 2 +3x 3 5 is valid for P. Then, in all roots, exactly one of the variables x 1 and x 3 must be odd. Thus, the given inequality is an XOR gadget, and it is associated with the pair {1, 3}. 7

8 3.2 Cutting planes Now we show how to use gadgets to derive cutting planes. Proposition 1 (Gadget Dicycle Inequalities) Consider an arbitrary collection of gadgets, of the form α r x β r for r R. Construct a directed multigraph, say G +, as follows. For i = 1,..., n and p {0, 1}, we have a node v p i. This node represents the statement x i Z p. There are directed arcs for each gadget in R and each pair {i, j} associated with that gadget (e.g., if all roots of the gadget with x i odd also have x j even, then we have an arc from vi 1 to vj 0, and another arc from vj 1 to v0 i ). Suppose there is an index i {1,..., n} such that there is a dicycle in G + passing through both vi 0 and vi 1. Let r(1),..., r(t) be the gadgets that correspond to the arcs in the dicycle. Then the gadget dicycle (GD) inequality t α r(s) x s=1 t β r(s) 1 s=1 is valid for P. Proof. Suppose that a solution x S were a root of all t gadgets simultaneously. The presence of a dipath in G + from vi 0 to vi 1 implies that, if x i is even, then it must be odd. Similarly, the dipath from vi 1 to v0 i implies that, if x i is odd, then it must be even. Since this yields a contradiction, at least one of the t gadgets must have a positive slack. We will see in the following two sections that, for the SSP and CPP, the number of facet-defining GD inequalities can be exponential in n. On the other hand, GD inequalities can sometimes be strengthened. In particular, it is possible that two or more arcs in the dicycle represent the same gadget. In this case, one can strengthen the GD inequality by including just one copy of each distinct gadget in the summation. 3.3 Exact separation We now show that the separation problem for the GD inequalities can be solved exactly in polynomial time. Proposition 2 (Separation of GD Inequalities) If R is a collection of gadgets whose size is bounded by a polynomial in n, then the separation problem for the corresponding GD inequalities can be solved in polynomial time. 8

9 Proof. Let x Q n + be the point to be separated. If any of the gadgets themselves are violated, then we immediately have a cutting plane. So assume that x satisfies all of the given gadgets. Construct the digraph G +, and set the weight of each arc to the slack of the associated gadget. Note that all of the arcs have non-negative weight. By construction, there is a oneto-one correspondence between violated GD inequalities and dicycles that have weight less than 1 and pass through some node vi 0 and its complement vi 1. The problem of finding a minimum-weight dicycle passing through a given pair can be reduced to two shortest (s, t)-dipath problems. So the separation problem can be reduced to 2n shortest-dipath calls in G +. Note that, if the digraph G + has parallel arcs between a pair of nodes, we can omit all apart from the one with smallest weight. So we can assume that G + contains O ( n 2) arcs. We can therefore solve each shortest-dipath subproblem in O ( n 2) time. Thus, the time taken to compute the minimumweight dicycle is O ( n 3). For a given combinatorial optimisation problem and a given family of gadgets, however, the total separation time could be larger or smaller, depending on the precise number of gadgets, the number of pairs associated with each gadget, and the time taken to compute the slacks of the gadgets. Note that any dicycle with weight less than 1 represents a violated inequality. As a result, the separation algorithm can potentially find up to n violated inequalities in a single call. Moreover, whenever a violated (or even near-violated) inequality has been found, one should check whether two or more arcs in the dicycle represent the same gadget. If so, then the inequality can be strengthened as mentioned in the previous subsection. 3.4 Heuristic separation for a special case We now consider the special case in which all of the gadgets in R are XOR and/or EQ gadgets. In this case, we can use an undirected multigraph G + instead of a directed one. For example, if an XOR gadget is associated with the pair {i, j}, all we need is an edge connecting vi 0 and vj 1 and an edge connecting vi 1 and vj 0. Then, every path in G+ linking some node vi 0 with its complement node vi 1 represents a GD inequality. We call these special GD inequalities gadget path (GP) inequalities. The use of an undirected graph in place of a directed one does not, in itself, lead to a significantly faster separation algorithm. On the other hand, it leads to a simple and fast separation heuristic. This heuristic is described in Algorithm 1. The following proposition shows that the heuristic can be implemented so that it runs efficiently. Proposition 3 Every step in Algorithm 1 from the computation of T onward can be implemented to take only O ( F 2) time. 9

10 Algorithm 1: Separation heuristic for GD inequalities. Input: An integer n 3, a collection R of XOR and EQ gadgets, and a fractional point x Q n + that satisfies all of the gadgets in R. Let F = { i {1,..., n} : 0 < x i < 1} denote the set of fractional variables; Construct the undirected version of G +, but include the nodes v 0 i and v 1 i only if i F ; for each remaining collection of parallel edges in G + do delete all apart from one of smallest weight (breaking ties arbitrarily); end Compute a minimum-weight spanning forest T in G + ; for each i F do if there is a path from v 0 i to v 1 i in T then if the weight of the path is less than 1 then output the corresponding violated GD inequality; end end end Proof. If we use a simple variant of Prim s algorithm [31], we can compute T in O ( F 2) time. There are then F candidates for the node vi 0. Consider a fixed candidate vi 0. Traverse T in a breadth-first manner, starting at v0 i. Along the way, record the total weight of the path in T from vi 0 to each node visited. This traversal of T takes O( F ) time. Once it is done, we can check in constant time if there is path from vi 0 to vi 1 with weight less than 1. So the time taken per candidate node is O( F ). In our preliminary computational experiments, this heuristic has proven to be remarkably effective. It is not clear to us whether the heuristic can be adapted to the (more general) GD inequalities. 3.5 Relationship with other approaches We now briefly relate our approach to two other general schemes for generating cuts, due to Caprara & Fischetti [3] and Borndoerfer & Weismantel [2]. Caprara & Fischetti [3] define a family of cuts, called {0, 1 2 }-cuts, for general ILPs. In Section 4 of their paper, they show that, if each constraint in the ILP has exactly two odd left-hand side coefficients, then the separation problem for {0, 1 2 }-cuts can be solved in polynomial time. Now, a constraint with two odd left-hand side coefficients is either an XOR gadget or an EQ gadget, depending on whether its right-hand side is odd or even (see 10

11 Example 2.) From this it can be shown that their result is a special case of ours. Moreover, our approach is more general, since (a) we do not impose restrictions on the parities of the left-hand side coefficients in the gadgets, and (b) we do not restrict ourselves to XOR and EQ gadgets only. Borndoerfer & Weismantel [2] present a related approach based on odd cycles in so-called conflict graphs. As far as we can tell, their approach is incomparable to ours. It is however shown in Section 4 of [23] that, if we are allowed to define new variables that are integer linear combinations of the original variables, then any cutting plane that can be derived using the conflict graph approach can also be derived using the above-mentioned approach of Caprara and Fischetti. Thus, if we allow such new variables, then our approach also generalises the conflict graph approach. 4 Application to the Stable Set Problem We now apply the gadget approach to the SSP. In Subsection 4.1, we define several kinds of gadget. In Subsection 4.2, we use our gadgets to obtain a new and exponentially large family of facet-defining inequalities. Then, in Subsection 4.3, we present separation algorithms and analyse their running time. 4.1 Gadgets for the SSP We know of many gadgets for the SSP. The ones that we have found most useful, in terms of deriving new valid inequalities for STAB(G), are all XOR and EQ gadgets. The following lemma introduces a trivial XOR gadget. Lemma 1 Every edge inequality (2) is an XOR gadget for STAB(G). Proof. In any root, exactly one of x i and x j must be equal to 1. The following theorem presents some more complex XOR gadgets. Theorem 1 Let i and j be a pair of non-adjacent nodes in G, let S be a subset of V \ {i, j}, and suppose that the inequality k S α kx k β defines a facet of STAB(G[S]), where the α k and β are all positive integers. Define { } β (i) = max α k x k : x STAB(G), x i = 1 k S { } β (j) = max α k x k : x STAB(G), x j = 1 k S { } β (i, j) = max α k x k : x STAB(G), x i = x j = 1, k S 11

12 and suppose that β (i, j) < β. Let = β (i) + β (j) β β (i, j). If = 1, then the inequality ( 2β 2β (i) + 1 ) x i + ( 2β 2β (j) + 1 ) x j + 2 α k x k 2β + 1 is an XOR gadget. If 2, then the inequality ( β β (i) + 1 ) x i + ( β β (j) + 1 ) x j + α k x k β + 1 is an XOR gadget. If 3, then the inequality ( β (j) β (i, j) 1 ) x i + ( β (i) β (i, j) 1 ) x j + α k x k β + 1 is also an XOR gadget. Proof. One can check that, if x i = x j = 0, all three inequalities reduce to k S α kx k β + ɛ for some positive ɛ. Thus, there are no roots with x i = x j = 0. One can also check that, if x i = x j = 1, the inequalities reduce to k S α kx k β (i, j) + ɛ for some positive ɛ. Thus, there are no roots with x i = x j = 1 either. If x i = 1 and x j = 0, the inequalities reduce to k S α kx k β (i). Therefore there exist roots with x i = 1 and x j = 0. A similar argument shows that there exist roots with x i = 0 and x j = 1. Here are two examples of XOR gadgets that can be derived in this way. Example 3 Let i and j be a pair of non-adjacent nodes in G, and let C V \ {i, j} be a clique such that (a) each node in C is adjacent to exactly one of i and j, (b) C n(i) 2 and (d) C n(j) 2. We have β = β (i) = β (j) = 1 and β (i, j) = 0. Thus, = 1 and the inequality k S k S k S x i + x j + 2 k C x k 3 (10) is an XOR gadget. Example 4 Let i and j be a pair of non-adjacent nodes in G, and let S V \ {i, j} be a set of 7 nodes such that G[S] is an odd hole. Suppose the nodes in S are labelled v(1),..., v(7), where nodes v(k) and v(k + 1) are adjacent for k = 1,..., 6. Suppose that node i is adjacent to nodes v(k) for k {1, 3, 5, 7}, and node j is adjacent to nodes v(k) for k {1, 2, 4, 6, 7}. We have β = β (i) = 3, β (j) = 2 and β (i, j) = 0. Thus, = 2 and the inequality x i + 2x j + k S x k 4 (11) is an XOR gadget. 12

13 Note that the gadget (11) does not have two odd left-hand side coefficients, as required in the method of Caprara and Fischetti. The following theorem presents some analogous EQ gadgets. Theorem 2 Let i, j, S, α, β, β (i), β (j), β (i, j) and be as in Theorem 1. If = 1, then the inequality ( 2β 2β (i) 1 ) x i + ( 2β 2β (j) 1 ) x j + 2 α k x k 2β is an EQ gadget. If = 2, then the inequality ( β β (i) 1 ) x i + ( β β (j) 1 ) x j + α k x k β is an EQ gadget. If 3, then the inequalities ( β β (i) 1 ) x i + ( β (i) β (i, j) + 1 ) x j + α k x k β ( β (j) β (i, j) + 1 ) x i + ( β β (j) 1 ) x j + α k x k β are EQ gadgets. Proof. If x i = x j = 0, all four inequalities reduce to k S α kx k β. Therefore there exist roots with x i = x j = 0. If x i = x j = 1, they all reduce to k S α kx k β (i, j). Therefore there also exist roots with x i = x j = 1. If x i = 1 and x j = 0, the inequalities reduce to k S α kx k β (i) + ɛ for some positive ɛ. Thus, there are no roots with x i = 1 and x j = 0. A similar argument shows that there are no roots with x i = 0 and x j = 1. Here are two examples of EQ gadgets that can be derived in this way. Example 5 Let i and j be a pair of non-adjacent nodes in G, and let C be a maximal subset of n(i) n(j) such that (a) C is a clique and (b) C 2. We have β = 1 and β (i) = β (j) = β (i, j) = 0. Thus, = 1 and the inequality x i + x j + 2 k C x k 2 (12) is an EQ gadget. Example 6 Let i and j be a pair of non-adjacent nodes in G, and let A be a subset of n(i) n(j) such that G[A] is an odd antihole. We have β = 2 and β (i) = β (j) = β (i, j) = 0. Thus, = 2 and the inequality k S k S k S k S x i + x j + k A x k 2 (13) is an EQ gadget. Again, the gadget (13) does not have two odd left-hand side coefficients. 13

14 Figure 4: Graphs for Examples 7 and Cutting planes for the SSP One can check that, if we use only the edge inequalities (2) as gadgets, the GP inequalities are precisely the odd hole inequalities. Thus, even with such a simple gadget, we obtain an exponentially large number of non-redundant GP inequalities. It turns out that, if we are allowed to use more complicated gadgets, we can obtain various lifted odd hole inequalities. Here are two examples. Example 7 Let G be the graph on the left of Fig. 4. Suppose we use the XOR gadget x 1 + x 7 1 and the following two EQ gadgets of type (12): x 1 + x ( x 2 + x 3 + x 8 ) 2 x 4 + x ( x 5 + x 6 + x 8 ) 2. One can check that, in G +, we have a path starting at v1 1, passing through v4 1 and v1 7, and ending at v0 1. This yields the GP inequality 2 7 x i + 4x 8 4. i=1 Dividing this by two, we obtain 7 i=1 x i + 2x 8 2. This is a lifting of the odd hole inequality x 1 + x 3 + x 4 + x 5 + x 7 2, and it can be shown to define a facet of STAB(G) (either by hand, or with the help of a software package such as PORTA [7]). Example 8 Let G be the graph on the right of Fig. 4. Suppose we use the following EQ gadget of type (12): x 1 + x ( x 2 + x 3 ) 2; and the following XOR gadget of type (10): x 1 + x x k 3. k=5

15 One can check that, in G +, the first gadget links v1 1 and v1 2, and the second links v2 1 with v0 1. This yields the GP inequality 2 8 x i 4. i=1 Dividing this by two, we obtain 8 i=1 x i 2. This is a lifting of the odd hole inequality x 1 + x 2 + x 4 + x 6 + x 7 2, and it can be shown to define a facet of STAB(G). One can check that the number of lifted odd hole inequalities obtainable in this way grows at least exponentially with n. It turns out that, with the aid of other gadgets, such as (11) and (13), we can obtain facet-defining GP inequalities that are not lifted odd hole inequalities. Example 9 Let G be the (unique) graph with 13 nodes and 31 edges such that: {1,..., 5} and {6,..., 10} induce 5-antiholes (which are also 5-holes); node 11 is adjacent to nodes 1,..., 10; node 12 is adjacent to nodes 1,..., 5; node 13 is adjacent to nodes 6,..., 10; nodes 12 and 13 are adjacent. (See Fig. 5.) Suppose we use the XOR gadget x 12 + x 13 1 and the following two EQ gadgets of type (13): x 11 + x 12 + x 11 + x x i 2 i=1 10 i=6 x i 2. One can check that, in G +, we have a path starting at v11 1, passing through v12 1 and v0 13, and ending at v0 11. This yields the GP inequality 2 ( ) 10 x 11 + x 12 + x 13 + x i 4. (14) This can be shown to define a facet of STAB(G). One can check that G does not contain any holes of cardinality 9, which means that the GP inequality is not a lifted odd hole inequality. i=1 15

16 11 {1,..., 5} {6,..., 10} Figure 5: Graph for Example 9. Readers who are familiar with the concept of Chvátal rank (see [8]) may find the following result of interest. Proposition 4 The GP inequality (14) has Chvátal rank larger than 1 with respect to the linear system defined by the clique and non-negativity inequalities. Proof. Since STAB(G) is full-dimensional and the inequality in question defines a facet, we can use the method in [20]. If we maximise the lefthand side of the inequality (14) subject to the clique and non-negativity inequalities, we obtain a fractional LP solution with x i equal to 0.25 for nodes 1,..., 10 and to 0.5 for nodes 11, 12 and 13. The profit of this solution is ( ) + (0.5 6) = 5.5. Since the right-hand side of (14) is less than 5.5, the inequality must have rank larger than Separation for the SSP Now we make some remarks on the separation problem for the GP inequalities for STAB(G). A first observation is that, if we include only the edge inequalities (2) as our gadgets, we effectively duplicate the exact odd hole separation routine of Gerards & Schrijver [14]. As mentioned in Subsection 2.3, this routine takes O ( V E + V 2 log V ) time. If we wish to add other gadgets in our collection, the running time increases. For example: If we also include our gadgets with S = 2, the time taken to compute the slacks of the gadgets increases to O ( E 2), and the total running time increases to O ( V 3 + E 2). If we also include our gadgets with S = 3, both the time to compute the slacks and the total running time increase to O ( V E 2). If we also include our gadgets with S = 4, both the time to compute the slacks and the total running time increase to O ( E 3). 16

17 In practice, rather than imposing restrictions on the cardinalities of S, it would be easier and faster to use greedy heuristics to find a good XOR gadget and EQ gadget for each pair i, j of non-adjacent nodes (where a good gadget means one with small slack). 5 Application to the Clique Partitioning Problem We now apply the gadget approach to the CPP. The structure of this section is identical to that of the previous one. 5.1 Gadgets for the CPP The following lemmas describe four simple, yet remarkably powerful, gadgets for the CPP. Lemma 2 For any {i, j} E n, the inequality x ij 0 (15) is a gadget for CPP n. This gadget is associated with n 2 distinct pairs of variables. Proof. Validity is trivial. If x is a root, then nodes i and j cannot lie in the same clique. This implies that, for any node k V \ {i, j}, the variables x ik and x jk cannot both be 1. There are n 2 candidates for k. Lemma 3 For any {i, j} E n, the inequality x ij 1 (16) is an EQ gadget for CPP n. This gadget too is associated with n 2 distinct pairs of variables. Proof. Validity is again trivial. If x is a root, then nodes i and j must lie in the same clique. This implies that, for any node k V \ {i, j}, the variables x ik and x jk must take the same value. Lemma 4 For any {i, j} E n and any k V n \ {i, j}, the transitivity inequality (5) is itself a gadget for the CPP. In particular, all roots satisfy the following implications: If x ij = 1 then x ik = 1 and x jk = 1; If x ik = 0 then x ij = 0 and x jk = 1; If x jk = 0 then x ij = 0 and x ik = 1. 17

18 Proof. One can check that there are two kinds of roots: either (a) x ij is 0 and exactly one of x ik and x jk are 1, or (b) all three variables are 1. In either case, the stated implications hold. Lemma 5 For any {i, j} E n and any k V n \ {i, j}, the inequality is an XOR gadget for CPP n. x ik + x jk 2x ij 1 (17) Proof. Validity follows from the fact that (17) is the sum of the transitivity inequality (5) and the inequality (15). One can check that in all roots, exactly one of the variables x ik and x jk is equal to one. Observe that the XOR gadget (17) is the sum of two other gadgets, namely, (5) and (15). This means that any valid inequality that can be derived using the four kinds of gadgets presented in this subsection can also be derived using only the first three kinds. Nevertheless, we will see that the XOR gadget is useful for the purposes of exposition, and also when devising separation algorithms. 5.2 Cutting planes for the CPP The following proposition shows that the XOR gadgets (17) alone are rather powerful. Proposition 5 Every GP inequality derived from the XOR gadgets (17) is equivalent to an OCW inequality (9) and vice-versa. Proof. Let G + be the graph obtained when R consists of the XOR gadgets. Note that there are two nodes in G + for each edge e E n, namely ve 0 and ve. 1 Moreover, G + is bipartite. Consider a path in G + that goes from one node to its complement node. Let c denote the number of edges in the path. Since G + is bipartite, c must be odd. Let e(1),..., e(c + 1) be the edges in E n that correspond to the first c nodes in the path, and note that the last node in the path corresponds to e(1). The ith edge in the path then corresponds to the XOR gadget x e(i) + x e(i+1) 2x f(i) 1, where f(i) is the edge having one end-node in common with e(i) and e(i + 1). The GP inequality can therefore be written as c ( ) xe(i) + x e(i+1) 2x f(i) c 1. i=1 Dividing this by two, we obtain the OCW inequality (9). Similarly, given any OCW inequality, one can construct a path in G + that yields a GP inequality that is twice the given OCW inequality. 18

19 Figure 6: Two facet-defining GP inequalities. It turns out that, if we use the EQ gadgets (16) and XOR gadgets (17) in combination, we can obtain facet-defining GP inequalities that are not OCW inequalities. Here are two examples. Example 10 Suppose we use the XOR gadgets x 12 + x 13 2x 23 1, x 13 + x 14 2x 34 1, x 14 + x 15 2x 45 1; and the EQ gadget x One can check that G + contains a path starting at the node v12 1, passing through v0 13, v1 14, v0 15, and ending at v0 12. This yields the GP inequality. x x x 14 + x 15 2x 23 + x 25 2x 34 2x (18) This can be shown to define a facet of CPP 5. Example 11 Suppose we use the XOR gadgets x 17 + x 27 2x 12 1, x 27 + x 37 2x 23 1, x 37 + x 47 2x 34 1, x 47 + x 57 2x 45 1, x 57 + x 67 2x 56 1, and the EQ gadget x One can check that G + contains a path starting at v17 1, passing through nodes v0 27,..., v0 67, and ending at v0 17. This yields the GP inequality x 17 +2x 27 +2x 37 +2x 47 +2x 57 +x 67 2x 12 2x 23 2x 34 2x 45 2x 56 5 (19) This can be shown to define a facet of CPP 7. The inequalities (18) and (19) are represented in Figure 6. Caprara & Fischetti [3] showed out that every OCW inequality has Chvátal rank 1. On the other hand, we have the following result. Proposition 6 The GP inequalities (18) and (19) have Chvátal rank larger than 1. 19

20 Figure 7: Basic fractional solutions to LP relaxations. Proof. Since CPP n is full-dimensional and the inequalities in question define facets of CPP 5 and CPP 7, respectively, we can again use the method in [20]. When n = 5, if we maximise the left-hand side of the inequality (18) subject to the transitivity and non-negativity inequalities, we obtain the fractional LP solution displayed on the left of Figure 7. (Here, the dotted lines represent variables with value 1/2.) The profit of this solution is 4. Since the right-hand side of (18) is less than 4, the inequality must have rank larger than 1. Similarly, when n = 7, if we maximise the left-hand side of the inequality (19) subject to the transitivity and non-negativity inequalities, we obtain the solution shown on the right of Figure 7. The profit of this solution is 6, whereas the right-hand side of (19) is less than 6. We leave it as an open question whether one can derive still more facetdefining inequalities by using a combination of the first three kinds of gadgets mentioned in the previous subsection. 5.3 Separation for the CPP Now we make some remarks on the separation problem for the GD and GP inequalities for the CPP. A major difference between the CPP and the SSP is that there are ( ) n 2 variables rather than V. Moreover, the number of gadgets described in Subsection 5.1 is O ( n 3). As a result, even if we use only those gadgets, the multigraph G + has O ( n 2) nodes and O ( n 3) edges. If we use the algorithm in [12], each shortest-path computation can be performed in O ( n 3) time. This leads to a total running time of O ( n 5). Since a running time of O ( n 5) is rather high, it may be preferable to use the GP separation heuristic described in Subsection 3.4. One can check that, for the CPP, the heuristic takes O ( f 2) time, where f is the number of fractional variables. We have performed some preliminary experiments with various cutting planes for the CPP. Our results indicate that, for typical CPP instances, the OW, OCW and GP inequalities tend to close around 83%, 88% and 92% 20

21 of the gap, respectively, between the LP upper bound and the optimum. However, we were using our heuristic to separate the GP inequalities, so the true gap closed by them may be higher. 6 Conclusions Up to now, gadgets have been used solely to prove hardness results. In this paper, we have shown that they can also be useful for deriving new cutting planes, along with accompanying separation algorithms. The results that we obtained with the stable set and clique partitioning problems suggest that this new approach has potential. A natural topic for future research is whether the gadget approach can be usefully applied to other N P-hard combinatorial optimisation problems. Problems worth consideration include, e.g., the TSP, the max-cut problem, the linear ordering problem [17] and the transitive acyclic subdigraph problem [26]. Another interesting topic is the possibility of defining gadgets that impose more complex structural properties on the roots, rather than merely imposing relationships between the parities of individual pairs of variables. References [1] B. Aspvall, M.F. Plass & R.E. Tarjan (1979) A linear-time algorithm for testing the truth of certain quantified boolean formulas. Inform. Proc. Lett., 8, [2] R. Borndörfer & R. Weismantel (2000) Set packing relaxations of integer programs. Math. Program., 88, [3] A. Caprara & M. Fischetti (1996) {0, 1 2 }-Chvátal-Gomory cuts. Math. Program., 74, [4] E. Cheng & W.H. Cunningham (1997) Wheel inequalities for stable set polytopes. Math. Program., 77, [5] E. Cheng & S. de Vries (2002). Antiweb-wheel inequalities and their separation problems over the stable set polytopes. Math. Program., 92, [6] S. Chopra & M.R. Rao (1993) The partition problem. Math. Program., 59, [7] T. Christof & A. Loebl, PORTA (polyhedron representation transformation algorithm). Software package, available for download at 21

22 [8] V. Chvátal (1973) Edmonds polytopes and a hierarchy of combinatorial problems. Discr. Math., 4, [9] M. Conforti, G. Cornuéjols & G. Zambelli (2014) Integer Programming. Cham, Switzerland: Springer. [10] S. Cook (1971) The complexity of theorem proving procedures. Proc. 3rd Ann. ACM Symp. Th. Comput. (STOC), pp [11] M. Deza, M. Grötschel & M. Laurent (1992) Clique-web facets for multicut polytopes. Math. Oper. Res., 17, [12] M.L. Fredman & R.E. Tarjan (1987) Fibonacci heaps and their uses in improved network optimization algorithms. J. of the ACM, 34, [13] M.R. Garey & D.S. Johnson (1979) Computers and Intractability: A Guide to the Theory of N P-Completeness. New York: Freeman. [14] A.H.M. Gerards & A.J. Schrijver (1986) Matrices with the Edmonds Johnson property. Combinatorica, 6, [15] M. Giandomenico & A.N. Letchford (2006) Exploring the relationship between max-cut and stable set relaxations. Math. Program., 106, [16] M. Grötschel, L. Lovász & A.J. Schrijver (1988) Stable sets in graphs. Chapter 9 of Geometric Algorithms in Combinatorial Optimization. New York: Wiley. [17] M. Grötschel, M. Jünger & G. Reinelt (1985) Facets of the linear ordering polytope. Math. Program., 33, [18] M. Grötschel & Y. Wakabayashi (1989) A cutting plane algorithm for a clustering problem. Math. Program., 45, [19] M. Grötschel & Y. Wakabayashi (1990) Facets of the clique partitioning polytope. Math. Program., 47, [20] M. Hartmann, M. Queyranne & Y. Wang (1999) On the Chvátal rank of certain inequalities. In G. Cornuéjols, R.E. Burkard & G.J. Woeginger (eds.) Proceedings of IPCO VII, pp Heidelberg: Springer. [21] J. Håstad (1999) Clique is hard to approximate within n 1 ɛ. Acta Math., 182, [22] M. Jünger et al. (eds.) 50 Years of Integer Programming: Berlin: Springer. [23] A.N. Letchford (2001) On disjunctive cuts for combinatorial optimization. J. Combin. Optim., 5,

23 [24] L. Lovász & A.J. Schrijver (1991) Cones of matrices and set-functions and 0-1 optimization. SIAM J. Optim., 1, [25] J.F. Marcotorchino (1981) Aggregation of Similarities in Automatic Classification (in French). Doctoral Thesis, Université Paris VI. [26] R. Müller (1996) On the partial order polytope of a digraph. Math. Program., 73, [27] G.L. Nemhauser & G. Sigismondi (1992) A strong cutting plane/branch-and-bound algorithm for node packing. J. Oper. Res. Soc., 43, [28] M. Oosten, J.H.G.C. Rutten & F.C.R. Spieksma (2001) The clique partitioning problem: facets and patching facets. Networks, 38, [29] M.W. Padberg (1973) On the facial structure of set packing polyhedra. Math. Program., 5, [30] M. Padberg & G. Rinaldi (1991) A branch-and-cut algorithm for the resolution of large-scale symmetric traveling salesman problems. SIAM Rev., 33, [31] R.C. Prim (1957) Shortest connection networks and some generalizations. Bell Sys. Tech. J., 36, [32] F. Rossi & S. Smriglio (2001) A branch-and-cut algorithm for the maximum cardinality stable set problem. Oper. Res. Lett., 28, [33] L. Trevisan, G.B. Sorkin, M. Sudan & D.P. Williamson (2000) Gadgets, approximation, and linear programming. SIAM J. Comput., 29, [34] Y. Wakabayashi (1986) Aggregation of Binary Relations: Algorithmic and Polyhedral Investigations. PhD Thesis, Augsburg University. 23

On Disjunctive Cuts for Combinatorial Optimization

On Disjunctive Cuts for Combinatorial Optimization Journal of Combinatorial Optimization, 5, 99 315, 001 c 001 Kluwer Academic Publishers. Manufactured in The Netherlands. On Disjunctive Cuts for Combinatorial Optimization ADAM N. LETCHFORD Department

More information

Projection Results for the k-partition Problem

Projection Results for the k-partition Problem Projection Results for the k-partition Problem Jamie Fairbrother Adam N. Letchford February 2016 Abstract The k-partition problem is an N P-hard combinatorial optimisation problem with many applications.

More information

Deciding Emptiness of the Gomory-Chvátal Closure is NP-Complete, Even for a Rational Polyhedron Containing No Integer Point

Deciding Emptiness of the Gomory-Chvátal Closure is NP-Complete, Even for a Rational Polyhedron Containing No Integer Point Deciding Emptiness of the Gomory-Chvátal Closure is NP-Complete, Even for a Rational Polyhedron Containing No Integer Point Gérard Cornuéjols 1 and Yanjun Li 2 1 Tepper School of Business, Carnegie Mellon

More information

Projection Results for the k-partition Problem

Projection Results for the k-partition Problem Projection Results for the k-partition Problem Jamie Fairbrother Adam N. Letchford To appear in Discrete Optimization Abstract The k-partition problem is an N P-hard combinatorial optimisation problem

More information

Facet-defining inequalities for the simple graph partitioning polytope

Facet-defining inequalities for the simple graph partitioning polytope Discrete Optimization 4 2007) 221 231 www.elsevier.com/locate/disopt Facet-defining inequalities for the simple graph partitioning polytope Michael M. Sørensen Aarhus School of Business, Department of

More information

On the Lovász Theta Function and Some Variants

On the Lovász Theta Function and Some Variants On the Lovász Theta Function and Some Variants Laura Galli Adam N. Letchford March 2017. To appear in Discrete Optimization. Abstract The Lovász theta function of a graph is a well-known upper bound on

More information

The Traveling Salesman Problem: Inequalities and Separation

The Traveling Salesman Problem: Inequalities and Separation The Traveling Salesman Problem: Inequalities and Separation Adam N. Letchford Department of Management Science, Lancaster University http://www.lancs.ac.uk/staff/letchfoa 1. The ILP formulation of the

More information

Separating Simple Domino Parity Inequalities

Separating Simple Domino Parity Inequalities Separating Simple Domino Parity Inequalities Lisa Fleischer Adam Letchford Andrea Lodi DRAFT: IPCO submission Abstract In IPCO 2002, Letchford and Lodi describe an algorithm for separating simple comb

More information

3.8 Strong valid inequalities

3.8 Strong valid inequalities 3.8 Strong valid inequalities By studying the problem structure, we can derive strong valid inequalities which lead to better approximations of the ideal formulation conv(x ) and hence to tighter bounds.

More information

Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs

Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs Franklin Djeumou Fomeni Konstantinos Kaparis Adam N. Letchford December 2013 Abstract The Reformulation-Linearization Technique (RLT),

More information

A 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 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 information

On the matrix-cut rank of polyhedra

On the matrix-cut rank of polyhedra On the matrix-cut rank of polyhedra William Cook and Sanjeeb Dash Computational and Applied Mathematics Rice University August 5, 00 Abstract Lovász and Schrijver (99) described a semi-definite operator

More information

On the separation of split cuts and related inequalities

On the separation of split cuts and related inequalities Math. Program., Ser. B 94: 279 294 (2003) Digital Object Identifier (DOI) 10.1007/s10107-002-0320-3 Alberto Caprara Adam N. Letchford On the separation of split cuts and related inequalities Received:

More information

On Matroid Parity and Matching Polytopes

On Matroid Parity and Matching Polytopes On Matroid Parity and Matching Polytopes Konstantinos Kaparis, Adam N. Letchford, Ioannis Mourtos October 017 Abstract The matroid parity problem is a natural extension of the matching problem to the matroid

More information

Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs

Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs Cutting Planes for RLT Relaxations of Mixed 0-1 Polynomial Programs Franklin Djeumou Fomeni Konstantinos Kaparis Adam N. Letchford To Appear in Mathematical Programming Abstract The Reformulation-Linearization

More information

When the Gomory-Chvátal Closure Coincides with the Integer Hull

When the Gomory-Chvátal Closure Coincides with the Integer Hull Date for Revised Manuscript: December 19, 2017 When the Gomory-Chvátal Closure Coincides with the Integer Hull Gérard Cornuéjols Yanjun Li Abstract Gomory-Chvátal cuts are prominent in integer programming.

More information

Lifting theorems and facet characterization for a class of clique partitioning inequalities

Lifting theorems and facet characterization for a class of clique partitioning inequalities Operations Research Letters 24 (1999) 235 243 www.elsevier.com/locate/orms Lifting theorems and facet characterization for a class of clique partitioning inequalities Hans-Jurgen Bandelt a, Maarten Oosten

More information

Polyhedral Approach to Integer Linear Programming. Tepper School of Business Carnegie Mellon University, Pittsburgh

Polyhedral 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 information

On the Polyhedral Structure of a Multi Item Production Planning Model with Setup Times

On the Polyhedral Structure of a Multi Item Production Planning Model with Setup Times CORE DISCUSSION PAPER 2000/52 On the Polyhedral Structure of a Multi Item Production Planning Model with Setup Times Andrew J. Miller 1, George L. Nemhauser 2, and Martin W.P. Savelsbergh 2 November 2000

More information

Travelling Salesman Problem

Travelling 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 information

The traveling salesman problem

The traveling salesman problem Chapter 58 The traveling salesman problem The traveling salesman problem (TSP) asks for a shortest Hamiltonian circuit in a graph. It belongs to the most seductive problems in combinatorial optimization,

More information

1.1 P, NP, and NP-complete

1.1 P, NP, and NP-complete CSC5160: Combinatorial Optimization and Approximation Algorithms Topic: Introduction to NP-complete Problems Date: 11/01/2008 Lecturer: Lap Chi Lau Scribe: Jerry Jilin Le This lecture gives a general introduction

More information

Multicommodity Flows and Column Generation

Multicommodity Flows and Column Generation Lecture Notes Multicommodity Flows and Column Generation Marc Pfetsch Zuse Institute Berlin pfetsch@zib.de last change: 2/8/2006 Technische Universität Berlin Fakultät II, Institut für Mathematik WS 2006/07

More information

Optimization Exercise Set n. 4 :

Optimization Exercise Set n. 4 : Optimization Exercise Set n. 4 : Prepared by S. Coniglio and E. Amaldi translated by O. Jabali 2018/2019 1 4.1 Airport location In air transportation, usually there is not a direct connection between every

More information

ON THE INTEGRALITY OF THE UNCAPACITATED FACILITY LOCATION POLYTOPE. 1. Introduction

ON THE INTEGRALITY OF THE UNCAPACITATED FACILITY LOCATION POLYTOPE. 1. Introduction ON THE INTEGRALITY OF THE UNCAPACITATED FACILITY LOCATION POLYTOPE MOURAD BAÏOU AND FRANCISCO BARAHONA Abstract We study a system of linear inequalities associated with the uncapacitated facility location

More information

My Favorite Theorem: Characterizations of Perfect Graphs

My Favorite Theorem: Characterizations of Perfect Graphs Konrad-Zuse-Zentrum für Informationstechnik Berlin Takustraße 7 D-14195 Berlin-Dahlem Germany MARTIN GRÖTSCHEL My Favorite Theorem: Characterizations of Perfect Graphs Preprint SC 99-17 (June 1999) MY

More information

Representations of All Solutions of Boolean Programming Problems

Representations of All Solutions of Boolean Programming Problems Representations of All Solutions of Boolean Programming Problems Utz-Uwe Haus and Carla Michini Institute for Operations Research Department of Mathematics ETH Zurich Rämistr. 101, 8092 Zürich, Switzerland

More information

The Maximum Flow Problem with Disjunctive Constraints

The Maximum Flow Problem with Disjunctive Constraints The Maximum Flow Problem with Disjunctive Constraints Ulrich Pferschy Joachim Schauer Abstract We study the maximum flow problem subject to binary disjunctive constraints in a directed graph: A negative

More information

Optimization Exercise Set n.5 :

Optimization Exercise Set n.5 : Optimization Exercise Set n.5 : Prepared by S. Coniglio translated by O. Jabali 2016/2017 1 5.1 Airport location In air transportation, usually there is not a direct connection between every pair of airports.

More information

On Non-Convex Quadratic Programming with Box Constraints

On Non-Convex Quadratic Programming with Box Constraints On Non-Convex Quadratic Programming with Box Constraints Samuel Burer Adam N. Letchford July 2008 Abstract Non-Convex Quadratic Programming with Box Constraints is a fundamental N P-hard global optimisation

More information

A Note on Representations of Linear Inequalities in Non-Convex Mixed-Integer Quadratic Programs

A Note on Representations of Linear Inequalities in Non-Convex Mixed-Integer Quadratic Programs A Note on Representations of Linear Inequalities in Non-Convex Mixed-Integer Quadratic Programs Adam N. Letchford Daniel J. Grainger To appear in Operations Research Letters Abstract In the literature

More information

Valid Inequalities and Separation for the Symmetric Sequential Ordering Problem

Valid Inequalities and Separation for the Symmetric Sequential Ordering Problem Valid Inequalities and Separation for the Symmetric Sequential Ordering Problem Adam N. Letchford Yanjun Li Draft, April 2014 Abstract The sequential ordering problem (SOP) is the generalisation of the

More information

Polynomial-Time Reductions

Polynomial-Time Reductions Reductions 1 Polynomial-Time Reductions Classify Problems According to Computational Requirements Q. Which problems will we be able to solve in practice? A working definition. [von Neumann 1953, Godel

More information

Solving the MWT. Recall the ILP for the MWT. We can obtain a solution to the MWT problem by solving the following ILP:

Solving the MWT. Recall the ILP for the MWT. We can obtain a solution to the MWT problem by solving the following ILP: Solving the MWT Recall the ILP for the MWT. We can obtain a solution to the MWT problem by solving the following ILP: max subject to e i E ω i x i e i C E x i {0, 1} x i C E 1 for all critical mixed cycles

More information

IBM Research Report. On a Connection between Facility Location and Perfect Graphs. Mourad Baïou University of Clermont

IBM Research Report. On a Connection between Facility Location and Perfect Graphs. Mourad Baïou University of Clermont RC24885 (W0910-175) October 28, 2009 Mathematics IBM Research Report On a Connection between Facility Location and Perfect Graphs Mourad Baïou University of Clermont Francisco Barahona IBM Research Division

More information

NP and Computational Intractability

NP and Computational Intractability NP and Computational Intractability 1 Polynomial-Time Reduction Desiderata'. Suppose we could solve X in polynomial-time. What else could we solve in polynomial time? don't confuse with reduces from Reduction.

More information

Lift-and-Project Inequalities

Lift-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 information

The extreme points of QSTAB(G) and its implications

The extreme points of QSTAB(G) and its implications Konrad-Zuse-Zentrum für Informationstechnik Berlin Takustraße 7 D-14195 Berlin-Dahlem Germany ARIE M.C.A. KOSTER ANNEGRET K. WAGLER The extreme points of QSTAB(G) and its implications ZIB-Report 06 30

More information

3.7 Cutting plane methods

3.7 Cutting plane methods 3.7 Cutting plane methods Generic ILP problem min{ c t x : x X = {x Z n + : Ax b} } with m n matrix A and n 1 vector b of rationals. According to Meyer s theorem: There exists an ideal formulation: conv(x

More information

Chapter 3. Complexity of algorithms

Chapter 3. Complexity of algorithms Chapter 3 Complexity of algorithms In this chapter, we see how problems may be classified according to their level of difficulty. Most problems that we consider in these notes are of general character,

More information

WHEN DOES THE POSITIVE SEMIDEFINITENESS CONSTRAINT HELP IN LIFTING PROCEDURES?

WHEN DOES THE POSITIVE SEMIDEFINITENESS CONSTRAINT HELP IN LIFTING PROCEDURES? MATHEMATICS OF OPERATIONS RESEARCH Vol. 6, No. 4, November 00, pp. 796 85 Printed in U.S.A. WHEN DOES THE POSITIVE SEMIDEFINITENESS CONSTRAINT HELP IN LIFTING PROCEDURES? MICHEL X. GOEMANS and LEVENT TUNÇEL

More information

Dual Consistent Systems of Linear Inequalities and Cardinality Constrained Polytopes. Satoru FUJISHIGE and Jens MASSBERG.

Dual Consistent Systems of Linear Inequalities and Cardinality Constrained Polytopes. Satoru FUJISHIGE and Jens MASSBERG. RIMS-1734 Dual Consistent Systems of Linear Inequalities and Cardinality Constrained Polytopes By Satoru FUJISHIGE and Jens MASSBERG December 2011 RESEARCH INSTITUTE FOR MATHEMATICAL SCIENCES KYOTO UNIVERSITY,

More information

On mixed-integer sets with two integer variables

On 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

Introduction to Mathematical Programming IE406. Lecture 21. Dr. Ted Ralphs

Introduction 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 information

Matroid Optimisation Problems with Nested Non-linear Monomials in the Objective Function

Matroid Optimisation Problems with Nested Non-linear Monomials in the Objective Function atroid Optimisation Problems with Nested Non-linear onomials in the Objective Function Anja Fischer Frank Fischer S. Thomas ccormick 14th arch 2016 Abstract Recently, Buchheim and Klein [4] suggested to

More information

LOVÁSZ-SCHRIJVER SDP-OPERATOR AND A SUPERCLASS OF NEAR-PERFECT GRAPHS

LOVÁSZ-SCHRIJVER SDP-OPERATOR AND A SUPERCLASS OF NEAR-PERFECT GRAPHS LOVÁSZ-SCHRIJVER SDP-OPERATOR AND A SUPERCLASS OF NEAR-PERFECT GRAPHS S. BIANCHI, M. ESCALANTE, G. NASINI, L. TUNÇEL Abstract. We study the Lovász-Schrijver SDP-operator applied to the fractional stable

More information

How tight is the corner relaxation? Insights gained from the stable set problem

How tight is the corner relaxation? Insights gained from the stable set problem How tight is the corner relaxation? Insights gained from the stable set problem Gérard Cornuéjols a,1, Carla Michini b,,, Giacomo Nannicini c,3 a Tepper School of Business, Carnegie Mellon University,

More information

constraints Ax+Gy» b (thus any valid inequalityforp is of the form u(ax+gy)» ub for u 2 R m + ). In [13], Gomory introduced a family of valid inequali

constraints Ax+Gy» b (thus any valid inequalityforp is of the form u(ax+gy)» ub for u 2 R m + ). In [13], Gomory introduced a family of valid inequali On the Rank of Mixed 0,1 Polyhedra Λ Gérard Cornuéjols Yanjun Li Graduate School of Industrial Administration Carnegie Mellon University, Pittsburgh, USA (corresponding author: gc0v@andrew.cmu.edu) February

More information

A packing integer program arising in two-layer network design

A packing integer program arising in two-layer network design A packing integer program arising in two-layer network design Christian Raack Arie M.C.A Koster Zuse Institute Berlin Takustr. 7, D-14195 Berlin Centre for Discrete Mathematics and its Applications (DIMAP)

More information

Breaking the Rectangle Bound Barrier against Formula Size Lower Bounds

Breaking the Rectangle Bound Barrier against Formula Size Lower Bounds Breaking the Rectangle Bound Barrier against Formula Size Lower Bounds Kenya Ueno The Young Researcher Development Center and Graduate School of Informatics, Kyoto University kenya@kuis.kyoto-u.ac.jp Abstract.

More information

3.7 Strong valid inequalities for structured ILP problems

3.7 Strong valid inequalities for structured ILP problems 3.7 Strong valid inequalities for structured ILP problems By studying the problem structure, we can derive strong valid inequalities yielding better approximations of conv(x ) and hence tighter bounds.

More information

Fast algorithms for even/odd minimum cuts and generalizations

Fast algorithms for even/odd minimum cuts and generalizations Fast algorithms for even/odd minimum cuts and generalizations András A. Benczúr 1 and Ottilia Fülöp 2 {benczur,otti}@cs.elte.hu 1 Computer and Automation Institute, Hungarian Academy of Sciences, and Department

More information

SEQUENTIAL AND SIMULTANEOUS LIFTING IN THE NODE PACKING POLYHEDRON JEFFREY WILLIAM PAVELKA. B.S., Kansas State University, 2011

SEQUENTIAL AND SIMULTANEOUS LIFTING IN THE NODE PACKING POLYHEDRON JEFFREY WILLIAM PAVELKA. B.S., Kansas State University, 2011 SEQUENTIAL AND SIMULTANEOUS LIFTING IN THE NODE PACKING POLYHEDRON by JEFFREY WILLIAM PAVELKA B.S., Kansas State University, 2011 A THESIS Submitted in partial fulfillment of the requirements for the degree

More information

Chapter 34: NP-Completeness

Chapter 34: NP-Completeness Graph Algorithms - Spring 2011 Set 17. Lecturer: Huilan Chang Reference: Cormen, Leiserson, Rivest, and Stein, Introduction to Algorithms, 2nd Edition, The MIT Press. Chapter 34: NP-Completeness 2. Polynomial-time

More information

Cutting planes from extended LP formulations

Cutting 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 information

Graph Coloring Inequalities from All-different Systems

Graph Coloring Inequalities from All-different Systems Constraints manuscript No (will be inserted by the editor) Graph Coloring Inequalities from All-different Systems David Bergman J N Hooker Received: date / Accepted: date Abstract We explore the idea of

More information

Cutting Plane Separators in SCIP

Cutting Plane Separators in SCIP Cutting Plane Separators in SCIP Kati Wolter Zuse Institute Berlin DFG Research Center MATHEON Mathematics for key technologies 1 / 36 General Cutting Plane Method MIP min{c T x : x X MIP }, X MIP := {x

More information

A Characterization of Graphs with Fractional Total Chromatic Number Equal to + 2

A Characterization of Graphs with Fractional Total Chromatic Number Equal to + 2 A Characterization of Graphs with Fractional Total Chromatic Number Equal to + Takehiro Ito a, William. S. Kennedy b, Bruce A. Reed c a Graduate School of Information Sciences, Tohoku University, Aoba-yama

More information

Coloring square-free Berge graphs

Coloring square-free Berge graphs Coloring square-free Berge graphs Maria Chudnovsky Irene Lo Frédéric Maffray Nicolas Trotignon Kristina Vušković September 30, 2015 Abstract We consider the class of Berge graphs that do not contain a

More information

On Some Polytopes Contained in the 0,1 Hypercube that Have a Small Chvátal Rank

On Some Polytopes Contained in the 0,1 Hypercube that Have a Small Chvátal Rank On ome Polytopes Contained in the 0,1 Hypercube that Have a mall Chvátal Rank Gérard Cornuéjols Dabeen Lee April 2016, revised July 2017 Abstract In this paper, we consider polytopes P that are contained

More information

3. Linear Programming and Polyhedral Combinatorics

3. Linear Programming and Polyhedral Combinatorics Massachusetts Institute of Technology 18.453: Combinatorial Optimization Michel X. Goemans April 5, 2017 3. Linear Programming and Polyhedral Combinatorics Summary of what was seen in the introductory

More information

Lower Bounds on the Sizes of Integer Programs Without Additional Variables

Lower Bounds on the Sizes of Integer Programs Without Additional Variables Lower Bounds on the Sizes of Integer Programs Without Additional Variables Volker Kaibel and Stefan Weltge Otto-von-Guericke-Universität Magdeburg, Germany {kaibel,weltge}@ovgu.de Abstract. For a given

More information

The Strength of Multi-row Aggregation Cuts for Sign-pattern Integer Programs

The Strength of Multi-row Aggregation Cuts for Sign-pattern Integer Programs The Strength of Multi-row Aggregation Cuts for Sign-pattern Integer Programs Santanu S. Dey 1, Andres Iroume 1, and Guanyi Wang 1 1 School of Industrial and Systems Engineering, Georgia Institute of Technology

More information

A New Facet Generating Procedure for the Stable Set Polytope

A New Facet Generating Procedure for the Stable Set Polytope 1 / 22 A New Facet Generating Procedure for the Stable Set Polytope Álinson S. Xavier a Manoel Campêlo b a Mestrado e Doutorado em Ciência da Computação Universidade Federal do Ceará Fortaleza, Brazil

More information

Matchings in hypergraphs of large minimum degree

Matchings in hypergraphs of large minimum degree Matchings in hypergraphs of large minimum degree Daniela Kühn Deryk Osthus Abstract It is well known that every bipartite graph with vertex classes of size n whose minimum degree is at least n/2 contains

More information

On Counting Lattice Points and Chvátal-Gomory Cutting Planes

On 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 information

FRACTIONAL PACKING OF T-JOINS. 1. Introduction

FRACTIONAL PACKING OF T-JOINS. 1. Introduction FRACTIONAL PACKING OF T-JOINS FRANCISCO BARAHONA Abstract Given a graph with nonnegative capacities on its edges, it is well known that the capacity of a minimum T -cut is equal to the value of a maximum

More information

arxiv: v1 [cs.cc] 5 Dec 2018

arxiv: v1 [cs.cc] 5 Dec 2018 Consistency for 0 1 Programming Danial Davarnia 1 and J. N. Hooker 2 1 Iowa state University davarnia@iastate.edu 2 Carnegie Mellon University jh38@andrew.cmu.edu arxiv:1812.02215v1 [cs.cc] 5 Dec 2018

More information

Paths and cycles in extended and decomposable digraphs

Paths and cycles in extended and decomposable digraphs Paths and cycles in extended and decomposable digraphs Jørgen Bang-Jensen Gregory Gutin Department of Mathematics and Computer Science Odense University, Denmark Abstract We consider digraphs called extended

More information

8. INTRACTABILITY I. Lecture slides by Kevin Wayne Copyright 2005 Pearson-Addison Wesley. Last updated on 2/6/18 2:16 AM

8. INTRACTABILITY I. Lecture slides by Kevin Wayne Copyright 2005 Pearson-Addison Wesley. Last updated on 2/6/18 2:16 AM 8. INTRACTABILITY I poly-time reductions packing and covering problems constraint satisfaction problems sequencing problems partitioning problems graph coloring numerical problems Lecture slides by Kevin

More information

MINI-TUTORIAL ON SEMI-ALGEBRAIC PROOF SYSTEMS. Albert Atserias Universitat Politècnica de Catalunya Barcelona

MINI-TUTORIAL ON SEMI-ALGEBRAIC PROOF SYSTEMS. Albert Atserias Universitat Politècnica de Catalunya Barcelona MINI-TUTORIAL ON SEMI-ALGEBRAIC PROOF SYSTEMS Albert Atserias Universitat Politècnica de Catalunya Barcelona Part I CONVEX POLYTOPES Convex polytopes as linear inequalities Polytope: P = {x R n : Ax b}

More information

Separation Techniques for Constrained Nonlinear 0 1 Programming

Separation 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 information

arxiv: v1 [math.co] 13 May 2016

arxiv: v1 [math.co] 13 May 2016 GENERALISED RAMSEY NUMBERS FOR TWO SETS OF CYCLES MIKAEL HANSSON arxiv:1605.04301v1 [math.co] 13 May 2016 Abstract. We determine several generalised Ramsey numbers for two sets Γ 1 and Γ 2 of cycles, in

More information

1 Matchings in Non-Bipartite Graphs

1 Matchings in Non-Bipartite Graphs CS 598CSC: Combinatorial Optimization Lecture date: Feb 9, 010 Instructor: Chandra Chekuri Scribe: Matthew Yancey 1 Matchings in Non-Bipartite Graphs We discuss matching in general undirected graphs. Given

More information

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Rani M. R, Mohith Jagalmohanan, R. Subashini Binary matrices having simultaneous consecutive

More information

The Complexity of Maximum. Matroid-Greedoid Intersection and. Weighted Greedoid Maximization

The Complexity of Maximum. Matroid-Greedoid Intersection and. Weighted Greedoid Maximization Department of Computer Science Series of Publications C Report C-2004-2 The Complexity of Maximum Matroid-Greedoid Intersection and Weighted Greedoid Maximization Taneli Mielikäinen Esko Ukkonen University

More information

AND ORBITOPES TIM JANUSCHOWSKI AND MARC E. PFETSCH

AND ORBITOPES TIM JANUSCHOWSKI AND MARC E. PFETSCH THE MAXIMUM k-colorable SUBGRAPH PROBLEM AND ORBITOPES TIM JANUSCHOWSKI AND MARC E. PFETSCH Abstract. Given an undirected node-weighted graph and a positive integer k, the maximum k-colorable subgraph

More information

Partial characterizations of clique-perfect graphs II: diamond-free and Helly circular-arc graphs

Partial characterizations of clique-perfect graphs II: diamond-free and Helly circular-arc graphs Partial characterizations of clique-perfect graphs II: diamond-free and Helly circular-arc graphs Flavia Bonomo a,1, Maria Chudnovsky b,2 and Guillermo Durán c,3 a Departamento de Matemática, Facultad

More information

New Integer Programming Formulations of the Generalized Travelling Salesman Problem

New Integer Programming Formulations of the Generalized Travelling Salesman Problem American Journal of Applied Sciences 4 (11): 932-937, 2007 ISSN 1546-9239 2007 Science Publications New Integer Programming Formulations of the Generalized Travelling Salesman Problem Petrica C. Pop Department

More information

A Polytope for a Product of Real Linear Functions in 0/1 Variables

A Polytope for a Product of Real Linear Functions in 0/1 Variables Don Coppersmith Oktay Günlük Jon Lee Janny Leung A Polytope for a Product of Real Linear Functions in 0/1 Variables Original: 29 September 1999 as IBM Research Report RC21568 Revised version: 30 November

More information

Minimal Valid Inequalities for Integer Constraints

Minimal Valid Inequalities for Integer Constraints Minimal Valid Inequalities for Integer Constraints Valentin Borozan LIF, Faculté des Sciences de Luminy, Université de Marseille, France borozan.valentin@gmail.com and Gérard Cornuéjols Tepper School of

More information

Lecture 1 : Probabilistic Method

Lecture 1 : Probabilistic Method IITM-CS6845: Theory Jan 04, 01 Lecturer: N.S.Narayanaswamy Lecture 1 : Probabilistic Method Scribe: R.Krithika The probabilistic method is a technique to deal with combinatorial problems by introducing

More information

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Michael M. Sørensen July 2016 Abstract Path-block-cycle inequalities are valid, and sometimes facet-defining,

More information

Finite-Domain Cuts for Graph Coloring

Finite-Domain Cuts for Graph Coloring Finite-Domain Cuts for Graph Coloring David Bergman J N Hooker April 01 Abstract We explore the idea of obtaining valid inequalities from a finite-domain formulation of a problem, rather than a 0-1 formulation

More information

CS675: Convex and Combinatorial Optimization Fall 2014 Combinatorial Problems as Linear Programs. Instructor: Shaddin Dughmi

CS675: Convex and Combinatorial Optimization Fall 2014 Combinatorial Problems as Linear Programs. Instructor: Shaddin Dughmi CS675: Convex and Combinatorial Optimization Fall 2014 Combinatorial Problems as Linear Programs Instructor: Shaddin Dughmi Outline 1 Introduction 2 Shortest Path 3 Algorithms for Single-Source Shortest

More information

Part II Strong lift-and-project cutting planes. Vienna, January 2012

Part II Strong lift-and-project cutting planes. Vienna, January 2012 Part II Strong lift-and-project cutting planes Vienna, January 2012 The Lovász and Schrijver M(K, K) Operator Let K be a given linear system in 0 1 variables. for any pair of inequalities αx β 0 and α

More information

Not Every GTSP Facet Induces an STSP Facet

Not Every GTSP Facet Induces an STSP Facet Not Every GTSP Facet Induces an STSP Facet Marcus Oswald, Gerhard Reinelt, and Dirk Oliver Theis Institute of Computer Science University of Heidelberg Im Neuenheimer Feld 38, 9120 Heidelberg, Germany

More information

Lecture notes on the ellipsoid algorithm

Lecture notes on the ellipsoid algorithm Massachusetts Institute of Technology Handout 1 18.433: Combinatorial Optimization May 14th, 007 Michel X. Goemans Lecture notes on the ellipsoid algorithm The simplex algorithm was the first algorithm

More information

The Multiple Traveling Salesman Problem with Time Windows: Bounds for the Minimum Number of Vehicles

The Multiple Traveling Salesman Problem with Time Windows: Bounds for the Minimum Number of Vehicles The Multiple Traveling Salesman Problem with Time Windows: Bounds for the Minimum Number of Vehicles Snežana Mitrović-Minić Ramesh Krishnamurti School of Computing Science, Simon Fraser University, Burnaby,

More information

3. Linear Programming and Polyhedral Combinatorics

3. Linear Programming and Polyhedral Combinatorics Massachusetts Institute of Technology 18.433: Combinatorial Optimization Michel X. Goemans February 28th, 2013 3. Linear Programming and Polyhedral Combinatorics Summary of what was seen in the introductory

More information

arxiv: v1 [cs.dm] 29 Oct 2012

arxiv: v1 [cs.dm] 29 Oct 2012 arxiv:1210.7684v1 [cs.dm] 29 Oct 2012 Square-Root Finding Problem In Graphs, A Complete Dichotomy Theorem. Babak Farzad 1 and Majid Karimi 2 Department of Mathematics Brock University, St. Catharines,

More information

NP-COMPLETE PROBLEMS. 1. Characterizing NP. Proof

NP-COMPLETE PROBLEMS. 1. Characterizing NP. Proof T-79.5103 / Autumn 2006 NP-complete problems 1 NP-COMPLETE PROBLEMS Characterizing NP Variants of satisfiability Graph-theoretic problems Coloring problems Sets and numbers Pseudopolynomial algorithms

More information

1 Perfect Matching and Matching Polytopes

1 Perfect Matching and Matching Polytopes CS 598CSC: Combinatorial Optimization Lecture date: /16/009 Instructor: Chandra Chekuri Scribe: Vivek Srikumar 1 Perfect Matching and Matching Polytopes Let G = (V, E be a graph. For a set E E, let χ E

More information

The Generalized Regenerator Location Problem

The Generalized Regenerator Location Problem The Generalized Regenerator Location Problem Si Chen Ivana Ljubić S. Raghavan College of Business and Public Affairs, Murray State University Murray, KY 42071, USA Faculty of Business, Economics, and Statistics,

More information

Cutting Planes in SCIP

Cutting Planes in SCIP Cutting Planes in SCIP Kati Wolter Zuse-Institute Berlin Department Optimization Berlin, 6th June 2007 Outline 1 Cutting Planes in SCIP 2 Cutting Planes for the 0-1 Knapsack Problem 2.1 Cover Cuts 2.2

More information

FINAL EXAM PRACTICE PROBLEMS CMSC 451 (Spring 2016)

FINAL EXAM PRACTICE PROBLEMS CMSC 451 (Spring 2016) FINAL EXAM PRACTICE PROBLEMS CMSC 451 (Spring 2016) The final exam will be on Thursday, May 12, from 8:00 10:00 am, at our regular class location (CSI 2117). It will be closed-book and closed-notes, except

More information

Approximation algorithms and hardness results for the clique packing problem. October, 2007

Approximation algorithms and hardness results for the clique packing problem. October, 2007 Approximation algorithms and hardness results for the clique packing problem F. Chataigner 1 G. Manić 2 Y.Wakabayashi 1 R. Yuster 3 1 Instituto de Matemática e Estatística Universidade de São Paulo, SP,

More information

The cocycle lattice of binary matroids

The cocycle lattice of binary matroids Published in: Europ. J. Comb. 14 (1993), 241 250. The cocycle lattice of binary matroids László Lovász Eötvös University, Budapest, Hungary, H-1088 Princeton University, Princeton, NJ 08544 Ákos Seress*

More information

Integer Programming ISE 418. Lecture 12. Dr. Ted Ralphs

Integer 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 information