## On weighted vs unweighted versions of combinatorial optimization problems (2001)

Venue: | Information and Computation |

Citations: | 18 - 2 self |

### BibTeX

@ARTICLE{Crescenzi01onweighted,

author = {Pierluigi Crescenzi and Riccardo Silvestri and Luca Trevisan},

title = {On weighted vs unweighted versions of combinatorial optimization problems},

journal = {Information and Computation},

year = {2001},

volume = {167},

pages = {2001}

}

### Years of Citing Articles

### OpenURL

### Abstract

We investigate the approximability properties of several weighted problems, by comparing them with the respective unweighted problems. For an appropriate (and very general) definition of niceness, we show that if a nice weighted problem is hard to approximate within r, then its polynomially bounded weighted version is hard to approximate within r − o(1). Then we turn our attention to specific problems, and we show that the unweighted versions of Min Vertex Cover, Min Sat, Max Cut, Max DiCut, Max 2Sat, and Max Exact kSat are exactly as hard to approximate as their weighted versions. We note in passing that Min Vertex Cover is exactly as hard to approximate as Min Sat. In order to prove the reductions for Max 2Sat, Max Cut, Max DiCut, and Max E3Sat we introduce the new notion of “mixing ” set and we give an explicit construction of such sets. These reductions give new non-approximability results for these problems. 1

### Citations

10885 | Computer and Intractability: A Guide to the Theory of NP-Completeness - Garey, Johnson - 1979 |

1686 |
The probabilistic method
- Alon, Spencer
- 1993
(Show Context)
Citation Context ...hosen independently and with probability pu. The expected measure of such a cut is clearly equal to � (u,v)∈E ω(u, v)(pu(1 − pv) + (1 − pu)pv). Using the method of conditional probabilities (see e.g. =-=[2]-=-), a cut with measure not smaller than this expectation can be constructed in polynomial time. Consider now the measure of Ĉ′ : for any edge (u, v) ∈ E we have to consider the edges connecting the Npu... |

706 | Proof verification and the hardness of approximation problems
- Arora, Lund, et al.
- 1998
(Show Context)
Citation Context ...on-approximability results for these problems. 1 Introduction One of the most important results obtained in the last decade in the field of computational complexity has been the so-called PCP-theorem =-=[3, 4]-=-, that is, the fact that any language in NP admits a probabilistic verifier of membership proofs using a logarithmic number of random bits and a constant number of queries. Besides being per se intere... |

640 | Some optimal inapproximability results
- H˚astad
(Show Context)
Citation Context ...f the PCP-theorem, several other refined proofs of the same result appeared in the literature in order to improve the performance of the verifier with respect to different query complexity parameters =-=[6, 15, 7, 5, 21, 22, 23]-=-. The last of these, due to Bellare, Goldreich and Sudan [5], and to H˚astad [23], allowed to significantly improve the lower bounds on the approximability of several important optimization problems, ... |

567 |
Optimization, approximation, and complexity classes
- Papadimitriou, Yannakakis
- 1991
(Show Context)
Citation Context ... the reduction of Bellare et al. does not preserve the existence of approximation algorithms with the same performance ratio. From a more structural point of view, we recall the following results. In =-=[31]-=- Papadimitriou and Yannakakis introduce the the class Max NP and study its extension to weighted problems. They also show that the very same algorithmic technique can be used to approximate both the p... |

359 | Probabilistic checking of proofs: A new characterization of NP
- Arora, Safra
- 1998
(Show Context)
Citation Context ...on-approximability results for these problems. 1 Introduction One of the most important results obtained in the last decade in the field of computational complexity has been the so-called PCP-theorem =-=[3, 4]-=-, that is, the fact that any language in NP admits a probabilistic verifier of membership proofs using a logarithmic number of random bits and a constant number of queries. Besides being per se intere... |

312 | P.; Ramanujan graphs - Lubotzky, Phillips, et al. - 1988 |

171 | Efficient probabilistically checkable proofs and applications to approximations - Bellare, Goldwasser, et al. - 1993 |

166 |
Clique is hard to approximate within n 1−ɛ
- Hastad
- 1999
(Show Context)
Citation Context ...f the PCP-theorem, several other refined proofs of the same result appeared in the literature in order to improve the performance of the verifier with respect to different query complexity parameters =-=[6, 15, 7, 5, 21, 22, 23]-=-. The last of these, due to Bellare, Goldreich and Sudan [5], and to H˚astad [23], allowed to significantly improve the lower bounds on the approximability of several important optimization problems, ... |

151 | Interactive proofs and the hardness of approximating cliques
- FEIGE, GOLDWASSER, et al.
- 1996
(Show Context)
Citation Context ...oved by using arguments similar to the reduction between Max Clique and zero-free bit PCP given by Bellare, Goldreich, and Sudan [5] and the reduction between PCP and Max Clique given by Feige et al. =-=[14]-=-, respectively. They have been independently proved by Marathe and Ravi [29]. Lemma 8 Min Vertex Cover ≤1 AP Min Sat. Proof: Let G = (V, E) be a graph. Fix an arbitrary total order ≤ in V (e.g. lexico... |

135 | Approximating maximum independent sets by excluding subgraphs
- Boppana, Halldorsson
- 1992
(Show Context)
Citation Context ... known whether the same performance ratio is obtainable. For instance, while the unweighted Max Clique problem is approximable within O(n/ log 2 n) where n denotes the number of vertices in the graph =-=[8]-=-, the best approximation algorithm for the weighted version of this problem reaches a factor of O((log log n) 2 n/ log 2 n) [20]. Finally, it is well-known that several NP-hard optimization problems t... |

116 | Improved non-approximability results
- Bellare, Sudan
- 1994
(Show Context)
Citation Context ...f the PCP-theorem, several other refined proofs of the same result appeared in the literature in order to improve the performance of the verifier with respect to different query complexity parameters =-=[6, 15, 7, 5, 21, 22, 23]-=-. The last of these, due to Bellare, Goldreich and Sudan [5], and to H˚astad [23], allowed to significantly improve the lower bounds on the approximability of several important optimization problems, ... |

116 | On Syntactic Versus Computational Views of Approximability
- Khanna, Motwani, et al.
- 1994
(Show Context)
Citation Context ...imable problem is reducible to a polynomially bounded weighted approximable problem by means of an approximation-preserving reduction (this result has been used by Khanna, Motwani, Sudan and Vazirani =-=[27]-=- to prove the APX-completeness of Max 3Sat). of Max Sat and of constraint satisfaction problems, instead, the repetition of clauses is typically allowed. However, when Max Cut for simple graphs is ref... |

116 |
Vertex packing: Structural properties and algorithms
- NEMHAUSER, TROTTER
- 1975
(Show Context)
Citation Context ...ching in the graph [18]. In order to obtain the first 2-approximate algorithm for the weighted version, instead, a linear programming formulation of the problem has been used by Nemhauser and Trotter =-=[30]-=-. In some cases, it is not known whether the same performance ratio is obtainable. For instance, while the unweighted Max Clique problem is approximable within O(n/ log 2 n) where n denotes the number... |

107 |
On the Approximation of Maximum Satisfiability
- Yannakakis
- 1994
(Show Context)
Citation Context ...tance x of size n, the performance ratio of the feasible solution T (x) with 3 Our results do not apply directly to Max 2Sat, but rather apply to its restricted version Max E2Sat. However, Yannakakis =-=[33]-=- proves that the approximation threshold of the two problems is the same. 3srespect to x is at most r(n). If a problem A admits an r-approximate polynomial-time algorithm for some constant r ≥ 1, then... |

94 | Introduction to the Theory of Complexity
- Bovet, Crescenzi
- 1993
(Show Context)
Citation Context ...blem is, given an instance x, to find an optimum solution y, that is, a solution whose measure is maximum or minimum depending on whether the problem is a maximization or a minimization one (see also =-=[9]-=-). In the following opt will denote the function that maps an instance x into the measure of an optimum solution. Let A be an optimization problem. For any instance x and for any solution y ∈ sol(x), ... |

93 |
D.S.: "Strong" NP-Completeness Results: Motivation, Examples, and Implications
- Garey, Johnson
- 1978
(Show Context)
Citation Context ... n) 2 n/ log 2 n) [20]. Finally, it is well-known that several NP-hard optimization problems turn out to be tractable whenever a polynomial bound is imposed on the weights that appear in the instance =-=[16]-=-. For example, the Min Partition problem can be solved in time O(nb) where n denotes the number of elements and b denotes the sum of their weights: clearly, if any upper bound were imposed on the weig... |

74 | Completeness in approximation classes
- Crescenzi, Panconesi
- 1991
(Show Context)
Citation Context ...l-time algorithm but A does not admit an (rA − ɛ)-approximate polynomial-time algorithm. Let A and B be two optimization problems and let α be a positive constant. A is said to be α-AP reducible to B =-=[11]-=-, in symbols A ≤α AP B, if two functions f and g exist such that: 1. For every x ∈ IA and for every r > 1, f(x, r) ∈ IB. 2. For every x ∈ IA, for every r > 1, and for every y ∈ solB(f(x, r)), g(x, y, ... |

47 |
Two prover protocols: low error at affordable rates
- Feige, Kilian
- 1994
(Show Context)
Citation Context |

42 | Gadgets, approximation and linear programming
- Trevisan, Sorkin, et al.
- 1996
(Show Context)
Citation Context ...ve thus shown that Max Weight Cut ≤α AP Max Cut. ⋄ 11 ⋄sRemark 17 A similar argument can be applied to Max DiCut thus showing that Max Weight DiCut ≤α AP Max DiCut, for any α > 1. From the results of =-=[5, 32, 23]-=- we have that, for any γ > 0 it is NP-hard to approximate Max Weight Cut (resp. Max DiCut) within 17/16−γ (resp. 13/12−γ). From the previous theorem we can extend these results to the unweighted versi... |

35 |
Testing of the long code and hardness for clique
- H˚astad
- 1996
(Show Context)
Citation Context |

28 | Hardness of approximating MAX k-CUT and its dual
- Kann, Khanna, et al.
- 1997
(Show Context)
Citation Context ...onti [1] and by Kann, Lagergren and Panconesi [26] and the Max k Cut (also known as Max k-Colorable Subgraph) studied by Papadimitriou and Yannakakis [31] and by Kann, Khanna, Lagergren and Panconesi =-=[25]-=-. We believe that our reduction can be extended to the general Max Sat problem. This should require the construction of k-ary (n, w, ɛ, δ)-mixing sets with k = n Ω(1) and a 1-AP reduction from the gen... |

25 | On approximation scheme preserving reducibility and its applications
- Crescenzi, Trevisan
(Show Context)
Citation Context ...ogically defined weighted problems, allowing both positive and negative weights. It turns out that the hardness of approximation dramatically increases when negative weights are admitted. Finally, in =-=[12]-=- Crescenzi and Trevisan proved that any arbitrarily weighted approximable problem is reducible to a polynomially bounded weighted approximable problem by means of an approximation-preserving reduction... |

21 | Multiplicative Number Theory (Second Edition - Davenport - 1980 |

18 |
Fast approximation for the knapsack and the sum of subset problems
- Ibarra, Kim
- 1975
(Show Context)
Citation Context ...p be a polynomial. We denote by A p the restriction of A to instances x such that the sum of the weights is at most p(|x|). The following result exploits the scaling techniques used by Ibarra and Kim =-=[24]-=- to develop a fully polynomial-time approximation scheme for the knapsack problem. Although the proof is only an application of very standard ideas, we give it for completeness. Theorem 4 Let A be a n... |

15 | Approximability of maximum splitting of k-sets and some other APX-complete problems
- Kann, Lagergren, et al.
- 1996
(Show Context)
Citation Context ... problems. Our results hold for any constraint satisfaction problem without unary constraints, such as the d-Regular Hypergraph Trasversal studied by Alimonti [1] and by Kann, Lagergren and Panconesi =-=[26]-=- and the Max k Cut (also known as Max k-Colorable Subgraph) studied by Papadimitriou and Yannakakis [31] and by Kann, Khanna, Lagergren and Panconesi [25]. We believe that our reduction can be extende... |

13 | Approximations via partitioning
- Halldorsson
- 1995
(Show Context)
Citation Context ...within O(n/ log 2 n) where n denotes the number of vertices in the graph [8], the best approximation algorithm for the weighted version of this problem reaches a factor of O((log log n) 2 n/ log 2 n) =-=[20]-=-. Finally, it is well-known that several NP-hard optimization problems turn out to be tractable whenever a polynomial bound is imposed on the weights that appear in the instance [16]. For example, the... |

11 | On Approximation algorithms for the minimum satisfiability problem
- Marathe, Ravi
- 1996
(Show Context)
Citation Context ...-free bit PCP given by Bellare, Goldreich, and Sudan [5] and the reduction between PCP and Max Clique given by Feige et al. [14], respectively. They have been independently proved by Marathe and Ravi =-=[29]-=-. Lemma 8 Min Vertex Cover ≤1 AP Min Sat. Proof: Let G = (V, E) be a graph. Fix an arbitrary total order ≤ in V (e.g. lexicographic order). We define an instance of Min Sat as follows. There is a vari... |

8 |
Non-oblivious local search for graph and hypergraph coloring problems
- Alimonti
- 1995
(Show Context)
Citation Context ...cit non-approximability results for these problems. Our results hold for any constraint satisfaction problem without unary constraints, such as the d-Regular Hypergraph Trasversal studied by Alimonti =-=[1]-=- and by Kann, Lagergren and Panconesi [26] and the Max k Cut (also known as Max k-Colorable Subgraph) studied by Papadimitriou and Yannakakis [31] and by Kann, Khanna, Lagergren and Panconesi [25]. We... |

8 |
bits, PCPs and nonapproximability
- Bellare, Goldreich, et al.
- 1995
(Show Context)
Citation Context |

5 | Weighted NP optimization problems: Logical definability and approximation properties
- Zimand
- 1999
(Show Context)
Citation Context ...eighted problems. They also show that the very same algorithmic technique can be used to approximate both the problems in Max NP and those in the weighted version of this class. More recently, Zimand =-=[34]-=- studied the approximability properties of logically defined weighted problems, allowing both positive and negative weights. It turns out that the hardness of approximation dramatically increases when... |

2 |
Tiny families of mixing functions
- Goldreich, Wigderson
- 1994
(Show Context)
Citation Context ... to derandomize the reduction. We first define a kind of graphs that will play the role of the random sets S w . These graphs are related to the mixing properties of expander graphs (see for instance =-=[19]-=-). Definition 12 Let w be any positive integer and let ɛ and δ be any two positive real numbers. A bipartite graph G = (V1, V2, E) is said to be (n, w, δ, ɛ)-mixing if |V1| = |V2| = n, and for any two... |

1 |
A compendium of NP optimization problems. Latest version available on-line at http://www.nada.kth.se/∼viggo/problemlist/compendium.html
- Crescenzi, Kann
(Show Context)
Citation Context ...any optimization problems have been studied both in the weighted and in the unweighted version. A list of such results is contained in the compendium of NP optimization problems by Crescenzi and Kann =-=[10]-=-. Bellare, Goldreich and Sudan [5] present a reduction from the Max Cut problem in multi-graphs (or, equivalently, in weighted graph with a polynomial bound on the weights) to Max Cut in simple graphs... |