## Zero Knowledge and the Chromatic Number (1996)

Venue: | Journal of Computer and System Sciences |

Citations: | 176 - 8 self |

### BibTeX

@INPROCEEDINGS{Feige96zeroknowledge,

author = {Uriel Feige and Joe Kilian},

title = {Zero Knowledge and the Chromatic Number},

booktitle = {Journal of Computer and System Sciences},

year = {1996},

pages = {278--287}

}

### Years of Citing Articles

### OpenURL

### Abstract

We present a new technique, inspired by zero-knowledge proof systems, for proving lower bounds on approximating the chromatic number of a graph. To illustrate this technique we present simple reductions from max-3-coloring and max-3-sat, showing that it is hard to approximate the chromatic number within \Omega\Gamma N ffi ), for some ffi ? 0. We then apply our technique in conjunction with the probabilistically checkable proofs of Hastad, and show that it is hard to approximate the chromatic number to within\Omega\Gamma N 1\Gammaffl ) for any ffl ? 0, assuming NP 6` ZPP. Here, ZPP denotes the class of languages decidable by a random expected polynomial-time algorithm that makes no errors. Our result matches (up to low order terms) the known gap for approximating the size of the largest independent set. Previous O(N ffi ) gaps for approximating the chromatic number (such as those by Lund and Yannakakis, and by Furer) did not match the gap for independent set, and do not extend...

### Citations

10922 |
Computers and Intractability: A Guide to the Theory of NP-Completeness
- Garey, Johnson
- 1979
(Show Context)
Citation Context ...ument we havesf (G)ff(G)sjV (G)j (2) Thus, an upper bound on ff(G) implies a corresponding lower bound onsf (G). 3.4 Randomized graph products Graph products amplify gaps in sizes of independent sets =-=[17]-=- and in chromatic number [26]. Randomized graph products were introduced by Berman and Schnitger [8] as a way of enhancing the effect of graph products. Applying the method of randomized graph product... |

1687 |
The probabilistic method
- Spencer
- 1992
(Show Context)
Citation Context ...ing a standard Chernoff bound. If PD (x i )s\Delta for all i, we can define the random variables Y i = (PD (x i )=\Delta)X i . Then, noting that jY i js1 and that S = \Delta P N i=1 Y i , we can use (=-=[2]-=-, Theorem A.16, p. 240) to conclude that P r(S ? \Delta) ! e \Gamma 2 =2N fors? 0. We can set \Delta = 2 \Gamma5n=6 by the condition of the lemma; settings= 2 2n=3 proves the conclusion of the lemma. ... |

1429 | Reducibility among combinatorial problems - Karp - 1972 |

382 |
On the hardness of approximating minimization problems
- Lund, Yannakakis
- 1993
(Show Context)
Citation Context ... assumption that NP 6` ZPP. After a long line of work [13, 4, 3, 6, 14, 7, 5, 21, 22], it has been shown hard to approximate ff(G) to within N 1\Gammaffl for any constant ffl ? 0. Lund and Yannakakis =-=[28]-=- reduced approximating ff(G) to approximating (G), thereby showing it hard to approximate (G) to within N c for some constant c ? 0. Subsequent work [24, 7, 16] has somewhat simplified and improved th... |

379 | Proofs that Yield Nothing But Their Validity or All languages in N P Have Zero-Knowledge Proof Systems
- Goldreich, Micali, et al.
- 1991
(Show Context)
Citation Context ...ating (G) within a factor of O(N ffl ), for some ffl ? 0. 4 The first step of the reduction uses standard techniques for constructing zero knowledge proof systems (specifically, we use protocols from =-=[19, 14]-=-). The second step is identical to the usual reduction to ff(G 0 ) [13]. The fact that the same reduction now also applies to (G 0 ) is a consequence of the weak duality relation between ff(G) and (G)... |

363 | Probabilistic Checking of Proofs: A New Characterization of NP
- Arora, Safra
- 1998
(Show Context)
Citation Context ... prove hardness of approximation results for ff(G) under the assumption that P 6= NP, and even sharper hardness of approximation results under the assumption that NP 6` ZPP. After a long line of work =-=[13, 4, 3, 6, 14, 7, 5, 21, 22]-=-, it has been shown hard to approximate ff(G) to within N 1\Gammaffl for any constant ffl ? 0. Lund and Yannakakis [28] reduced approximating ff(G) to approximating (G), thereby showing it hard to app... |

318 | A Parallel Repetition Theorem - Raz - 1998 |

262 |
On the Ratio of the Optimal Integral and Fractional Covers
- Lovász
- 1975
(Show Context)
Citation Context ...stribution I on G's independent sets, choosing I / I covers any v 2 V (G) with probability at least 1=k. (This is not the standard definition forsf (G), but is equivalent.) Clearly,sf (G)s(G). Lovasz =-=[27]-=- has shown that for any graph G, (G) 1 + ln ff(G)sf (G)s(G): (1) 6 Thus,sf (G) approximates (G) to within a logarithmic factor, so strong hardness results for approximatingsf (G) apply to (G) as well.... |

208 |
Proof verification and intractability of approximation problems
- Arora, Lund, et al.
- 1998
(Show Context)
Citation Context ... prove hardness of approximation results for ff(G) under the assumption that P 6= NP, and even sharper hardness of approximation results under the assumption that NP 6` ZPP. After a long line of work =-=[13, 4, 3, 6, 14, 7, 5, 21, 22]-=-, it has been shown hard to approximate ff(G) to within N 1\Gammaffl for any constant ffl ? 0. Lund and Yannakakis [28] reduced approximating ff(G) to approximating (G), thereby showing it hard to app... |

205 | Free bits, PCPs, and nonapproximability— towards tight results
- Bellare, Goldreich, et al.
- 1998
(Show Context)
Citation Context ... prove hardness of approximation results for ff(G) under the assumption that P 6= NP, and even sharper hardness of approximation results under the assumption that NP 6` ZPP. After a long line of work =-=[13, 4, 3, 6, 14, 7, 5, 21, 22]-=-, it has been shown hard to approximate ff(G) to within N 1\Gammaffl for any constant ffl ? 0. Lund and Yannakakis [28] reduced approximating ff(G) to approximating (G), thereby showing it hard to app... |

155 | Interactive proofs and the hardness of approximating cliques
- Feige, Goldwasser, et al.
- 1996
(Show Context)
Citation Context ...r to provide evidence that approximating the above quantities is hard. An early result of Garey and Johnson [18] showed it NP-hard to approximates(G) to within any constant less than 2. More recently =-=[13]-=-, a connection was established between the hardness of approximating ff(G) and the construction of probabilistically checkable proofs (PCPs). Using this connection it became possible to prove hardness... |

135 | Approximating maximum independent sets by excluding subgraphs - Boppana, Halldorsson - 1992 |

117 | Improved non-approximability results
- Bellare, Sudan
- 1994
(Show Context)
Citation Context |

110 | Simulating BPP using a general weak random source
- Zuckerman
- 1996
(Show Context)
Citation Context ...t of approximating ff(G) within a factor of O(N ffl ), for some ffl ? 0. The above reduction is randomized, due to its second step, and relies on the assumption that NP 6` BPP. It can be derandomized =-=[31, 1]-=-, though the value of ffl obtained by using known derandomization techniques is not as good as the one obtained by the randomized reduction. For this 3 reason, hardness of approximation results for ff... |

80 | The hardness of approximation: Gap location
- Petrank
- 1994
(Show Context)
Citation Context ...ductions to chromatic number In this section we give RPCPs that, using the methodology of the previous section, imply new reductions to chromatic number. 4.1 A reduction from max-3-coloring Petrank's =-=[29]-=- reduction from max-3-sat to max-3-coloring implies the following: For any NP-complete language L and input x one can produce a graph H such that 1. if x 2 L, then H is 3-colorable, and 2. if x 62 L a... |

73 | Derandomized Graph Products
- Alon, Feige, et al.
- 1995
(Show Context)
Citation Context ...t of approximating ff(G) within a factor of O(N ffl ), for some ffl ? 0. The above reduction is randomized, due to its second step, and relies on the assumption that NP 6` BPP. It can be derandomized =-=[31, 1]-=-, though the value of ffl obtained by using known derandomization techniques is not as good as the one obtained by the randomized reduction. For this 3 reason, hardness of approximation results for ff... |

73 |
On the Complexity of Approximating the Independent Set Problem
- Berman, Schnitger
- 1992
(Show Context)
Citation Context ...for randomizing long-code based proofs and prove Theorem 1. 2 Putting the new approach in perspective A relatively simple proof that ff(G) is hard to approximate follows ideas of Berman and Schnitger =-=[8]-=-. Simple hardness results for ff(G): 1. Reduce the problem of approximating max-3SAT to approximating ff(G 0 ) within a constant factor in a graph G 0 that has linear size independent sets. 2. Use a t... |

73 | A still better performance guarantee for approximate graph coloring
- Halldórsson
- 1993
(Show Context)
Citation Context ...time algorithms for approximating (G) and ff(G) do not guarantee a good approximation ratio. The best approximation ratios known to be efficiently achievable are O(N(log log N) 2 =(log N) 3 ) for (G) =-=[20]-=-, and O(N=(log N) 2 ) for ff(G) [9]. As well as computing an approximation for (G) or ff(G), these algorithms produce a coloring (or independent set) that is within the guaranteed ratio from optimal. ... |

71 | On the hardness of approximating the chromatic number
- Khanna, Linial, et al.
(Show Context)
Citation Context ... for any constant ffl ? 0. Lund and Yannakakis [28] reduced approximating ff(G) to approximating (G), thereby showing it hard to approximate (G) to within N c for some constant c ? 0. Subsequent work =-=[24, 7, 16]-=- has somewhat simplified and improved this reduction. Applying the most efficient reduction [16] to the strongest result known for independent set [22] implies that it is NP-hard (under randomized red... |

67 |
Clique is hard to approximate within n 1\Gammaffl
- Hastad
- 1999
(Show Context)
Citation Context ... ever picking an f i that is not low-biased by m X i=1 e \Gamma2 1 6 ns ism k X i=1 e \Gamma2 1 6 n i ; which implies the lemma. 2 16 5.3 Overview of Hastad's PCP. We review here the PCP described in =-=[22]-=-. Hastad starts with a redundant 3CNF formula F (pedantically, a family of formulas) with n variables and 2n clauses in which each variable x i appears exactly 6 times. The construction requires that ... |

57 |
The complexity of near optimal graph coloring
- Garey, Johnson
- 1976
(Show Context)
Citation Context ... within the guaranteed ratio from optimal. The theory of NP-completeness can be used in order to provide evidence that approximating the above quantities is hard. An early result of Garey and Johnson =-=[18]-=- showed it NP-hard to approximates(G) to within any constant less than 2. More recently [13], a connection was established between the hardness of approximating ff(G) and the construction of probabili... |

41 | Randomized graph products, chromatic numbers, and the Lovász ϑ-function
- Feige
- 1997
(Show Context)
Citation Context ...elation between ff(G) and (G), and the particular structure of zero knowledge PCPs. The third step uses a version of randomized graph products that is suitable for amplifying gaps in chromatic number =-=[12]-=-, rather than the randomized graph products that are used in the context of independent sets [8, 10]. To obtain stronger hardness of approximation results for (G), we start from a PCP with few amortiz... |

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

28 | Efficient multi-prover interactive proofs with applications to approximation problems - Bellare, Goldwasser, et al. - 1993 |

22 | Algorithms for Approximate Graph Coloring
- Blum
- 1991
(Show Context)
Citation Context ...ce the problem of approximating max-3SAT to approximating ff(G 0 ) within a constant factor in a graph G 0 that has linear size independent sets. 2. Use a technique known as randomized graph products =-=[8, 10]-=- to produce a new graph G that only has small independent sets. The problem of approximating ff(G 0 ) within a constant factor is reduced to that of approximating ff(G) within a factor of O(N ffl ), f... |

20 |
Improved hardness results for approximating the chromatic number
- Furer
- 1994
(Show Context)
Citation Context ... for any constant ffl ? 0. Lund and Yannakakis [28] reduced approximating ff(G) to approximating (G), thereby showing it hard to approximate (G) to within N c for some constant c ? 0. Subsequent work =-=[24, 7, 16]-=- has somewhat simplified and improved this reduction. Applying the most efficient reduction [16] to the strongest result known for independent set [22] implies that it is NP-hard (under randomized red... |

15 |
Graph products and chromatic numbers
- Linial, Vazirani
- 1989
(Show Context)
Citation Context ...G)j (2) Thus, an upper bound on ff(G) implies a corresponding lower bound onsf (G). 3.4 Randomized graph products Graph products amplify gaps in sizes of independent sets [17] and in chromatic number =-=[26]-=-. Randomized graph products were introduced by Berman and Schnitger [8] as a way of enhancing the effect of graph products. Applying the method of randomized graph products [8, 10] to the graph G 0 ob... |

11 |
Low communication 2-prover zero-knowledge proofs for np
- Dwork, Feige, et al.
- 1992
(Show Context)
Citation Context .... Our new approach for proving hardness results for approximating (G) does not provide a direct reduction from ff(G) to (G). Instead, it is based on randomized versions of PCPs (RPCPs), introduced in =-=[11]-=-. Zero-knowledge PCPs are a subfamily of randomized PCPs. Indeed, some of the randomized PCPs we use are in fact zero-knowledge PCPs. Similar PCPs were previously used in [25] to show the hardness of ... |

5 |
Two prover protocols–low error at affordable rates
- Feige, Kilian
(Show Context)
Citation Context |

3 |
Zero knowledge and the chromatic number (preliminary version
- Feige, Kilian
- 1996
(Show Context)
Citation Context ...earlier PCPs based on the long code (those of [5, 21]), as well as to simplified versions of the PCP of [22], recently announced by Hastad (private communication). In an earlier version of this paper =-=[15]-=-, we used a slightly different method to randomize the PCPs of [5, 21]. This method had the advantage of having the prover use only a constant number of random bits. The RPCPs that we constructed were... |

3 | On the Complexity of Statistical Reasoning
- Kilian, Naor
- 1995
(Show Context)
Citation Context ...s (RPCPs), introduced in [11]. Zero-knowledge PCPs are a subfamily of randomized PCPs. Indeed, some of the randomized PCPs we use are in fact zero-knowledge PCPs. Similar PCPs were previously used in =-=[25]-=- to show the hardness of certain probabilistic inference problems. In its simpler form, our reduction has the following structure: Simple hardness results for (G): 1. Reduce the problem of approximati... |