## Derandomization, witnesses for Boolean matrix multiplication and construction of perfect hash functions (1996)

### Cached

### Download Links

Venue: | Algorithmica |

Citations: | 61 - 5 self |

### BibTeX

@ARTICLE{Alon96derandomization,witnesses,

author = {Noga Alon and Moni Naor},

title = {Derandomization, witnesses for Boolean matrix multiplication and construction of perfect hash functions},

journal = {Algorithmica},

year = {1996},

volume = {16},

pages = {16--434}

}

### OpenURL

### Abstract

Small sample spaces with almost independent random variables are applied to design efficient sequential deterministic algorithms for two problems. The first algorithm, motivated by the attempt to design efficient algorithms for the All Pairs Shortest Path problem using fast matrix multiplication, solves the problem of computing witnesses for the Boolean product of two matrices. That is, if A and B are two n by n matrices, and C = AB is their Boolean product, the algorithm finds for every entry Cij = 1 a witness: an index k so that Aik = Bkj = 1. Its running time exceeds that of computing the product of two n by n matrices with small integer entries by a polylogarithmic factor. The second algorithm is a nearly linear time deterministic procedure for constructing a perfect hash function for a given n-subset of {1,..., m}.

### Citations

1710 | The Probabilistic Method - Alon, Spencer - 1992 |

830 |
Matrix multiplication via arithmetic progressions
- Coppersmith, Winograd
- 1990
(Show Context)
Citation Context ...= �n k=1(Aik ∧ Bkj). The n3 time method that evaluates these expressions gives for every i, j for which Cij = 1 all the k’s for which Aik = Bkj = 1. The subcubic methods on the other hand (see, e.g., =-=[8]-=-) consider A and B as matrices of integers and do not provide any of these k’s. We call a k such that Aik = Bkj = 1 a witness (for the fact that Cij = 1). We want to compute in addition to the matrix ... |

379 | A Simple Parallel Algorithm for the Maximal Independent Set Problem
- Luby
- 1986
(Show Context)
Citation Context ...oportional to the size of the space. The size of the space is usually some polynomial in the input size. Thus again this approach suffers from considerable increase in time. This approach is taken in =-=[17]-=-, [1], [15] using probability spaces that are k-wise independent, and in [5], [23] using small bias probability spaces and almost k-wise independence (see definition below in subsection 1.3). Our goal... |

282 |
Probabilistic construction of deterministic algorithms: Approximating packing integer programs
- Raghavan, Thompson
- 1988
(Show Context)
Citation Context ...hich usually takes time proportional to the input size. Thus, the cost of derandomization though polynomial, can considerably increase the complexity of the algorithm. This approach is taken in [29], =-=[25]-=-, (cf., also [6].) A different approach for finding a good point is to show that the random choices made need not be fully independent, i.e. even if some limited form of independence is obeyed, then t... |

269 | Simple construction of almost k-wise independent random variables. Random Struct
- Alon, Goldreich, et al.
- 1992
(Show Context)
Citation Context ...lynomial in the input size. Thus again this approach suffers from considerable increase in time. This approach is taken in [17], [1], [15] using probability spaces that are k-wise independent, and in =-=[5]-=-, [23] using small bias probability spaces and almost k-wise independence (see definition below in subsection 1.3). Our goal in this work is to use these methods without incurring a significant penalt... |

256 | Small-bias probability spaces: Efficient constructions and applications
- Naor, Naor
- 1993
(Show Context)
Citation Context ...ial in the input size. Thus again this approach suffers from considerable increase in time. This approach is taken in [17], [1], [15] using probability spaces that are k-wise independent, and in [5], =-=[23]-=- using small bias probability spaces and almost k-wise independence (see definition below in subsection 1.3). Our goal in this work is to use these methods without incurring a significant penalty in t... |

217 |
Storing a sparse table with O(1) worst case access time
- Fredman, Komlos, et al.
- 1984
(Show Context)
Citation Context ...od of computing h(x). • An efficient construction - given S there should be an efficient way of finding h ∈ H that is perfect for S Perfect hash functions have been investigated extensively (see e.g. =-=[9, 10, 11, 12, 16, 19, 21, 26, 27, 30]-=-). It is known (and not too difficult to show, see [11], [16], [24]) that the minimum possible number of bits required to represent such a mapping is Θ(n + log log m) for all m ≥ 2n. Fredman, Komlós a... |

216 |
A Fast and Simple Randomized Parallel Algorithm for the Maximal Independent Set Problem
- Alon, Babai, et al.
- 1986
(Show Context)
Citation Context ...onal to the size of the space. The size of the space is usually some polynomial in the input size. Thus again this approach suffers from considerable increase in time. This approach is taken in [17], =-=[1]-=-, [15] using probability spaces that are k-wise independent, and in [5], [23] using small bias probability spaces and almost k-wise independence (see definition below in subsection 1.3). Our goal in t... |

151 |
Data structures and algorithm 1: Sorting and Searching
- Mehlhorn
- 1984
(Show Context)
Citation Context ...od of computing h(x). • An efficient construction - given S there should be an efficient way of finding h ∈ H that is perfect for S Perfect hash functions have been investigated extensively (see e.g. =-=[9, 10, 11, 12, 16, 19, 21, 26, 27, 30]-=-). It is known (and not too difficult to show, see [11], [16], [24]) that the minimum possible number of bits required to represent such a mapping is Θ(n + log log m) for all m ≥ 2n. Fredman, Komlós a... |

115 | Construction of asymptotically good low-rate error-correcting codes through pseudo-random graphs
- Alon, Bruck, et al.
- 1992
(Show Context)
Citation Context .... Therefore if 1/ɛ is logarithmic in n and c is at most logarithmic in n the size of the probability space is still polylogarithmic in n. To be more precise, the construction of [23], as optimized in =-=[2]-=-, yields a probability space of size O( ) and the ones in [5] yield probability spaces of size O( c2 log2 n ɛ2 ). 4 i∈S c log n ɛ 3s2 Boolean matrix multiplication with witnesses All the matrices in t... |

82 | A fast parallel algorithm for the maximal independent set problem
- Karp, Wigderson
- 1984
(Show Context)
Citation Context ...to the size of the space. The size of the space is usually some polynomial in the input size. Thus again this approach suffers from considerable increase in time. This approach is taken in [17], [1], =-=[15]-=- using probability spaces that are k-wise independent, and in [5], [23] using small bias probability spaces and almost k-wise independence (see definition below in subsection 1.3). Our goal in this wo... |

71 | On the exponent of the All Pairs Shortest Path problem - Alon, Galil, et al. - 1991 |

62 |
Storing a sparse table
- Tarjan, Yao
- 1979
(Show Context)
Citation Context ...od of computing h(x). • An efficient construction - given S there should be an efficient way of finding h ∈ H that is perfect for S Perfect hash functions have been investigated extensively (see e.g. =-=[9, 10, 11, 12, 16, 19, 21, 26, 27, 30]-=-). It is known (and not too difficult to show, see [11], [16], [24]) that the minimum possible number of bits required to represent such a mapping is Θ(n + log log m) for all m ≥ 2n. Fredman, Komlós a... |

58 |
Ten lectures on the probabilistic method
- Spencer
- 1987
(Show Context)
Citation Context ...ator which usually takes time proportional to the input size. Thus, the cost of derandomization though polynomial, can considerably increase the complexity of the algorithm. This approach is taken in =-=[29]-=-, [25], (cf., also [6].) A different approach for finding a good point is to show that the random choices made need not be fully independent, i.e. even if some limited form of independence is obeyed, ... |

51 | Removing randomness in parallel computation without a processor penalty
- Luby
- 1988
(Show Context)
Citation Context ... since each probability space is small the total work is increased by only a polylogarithmic factor. We note that the two approaches described above had been combined in a different way previously in =-=[18, 7, 22]-=-. The random variables used there were k-wise independent resulting in a probability space of size O(n k ). This probability space was then searched using an estimator function in O(k log n) steps. Th... |

51 | The probabilistic method yields deterministic parallel algorithms
- Motwani, Naor, et al.
- 1989
(Show Context)
Citation Context ... since each probability space is small the total work is increased by only a polylogarithmic factor. We note that the two approaches described above had been combined in a different way previously in =-=[18, 7, 22]-=-. The random variables used there were k-wise independent resulting in a probability space of size O(n k ). This probability space was then searched using an estimator function in O(k log n) steps. Th... |

48 |
On the size of separating systems and families of perfect hash functions
- Fredman, Komlós
- 1984
(Show Context)
Citation Context |

43 |
The spatial complexity of oblivious k-probe hash functions
- Schmidt, Siegel
- 1990
(Show Context)
Citation Context |

41 | On the all pairs shortest path problem
- Seidel
- 1992
(Show Context)
Citation Context ...ent ω < 2.376 and is due to Coppersmith and Winograd [8]. For two functions f(n) and g(n) we let g(n) = Õ(f(n)) denote the statement that g(n) is O(f(n)(log n) O(1) ). Several researchers (see, e.g., =-=[28]-=-, [4]) observed that there is a simple randomized algorithm that computes witnesses in Õ(nω ) time. In Section 2 we describe a deterministic algorithm for computing the witnesses in Õ(nω ) time. It is... |

32 |
Fredman–Komlós bounds and information theory
- Körner
- 1986
(Show Context)
Citation Context |

30 |
On witnesses for Boolean matrix multiplications and for shortest paths
- Alon, Galil, et al.
(Show Context)
Citation Context ...< 2.376 and is due to Coppersmith and Winograd [8]. For two functions f(n) and g(n) we let g(n) = Õ(f(n)) denote the statement that g(n) is O(f(n)(log n) O(1) ). Several researchers (see, e.g., [28], =-=[4]-=-) observed that there is a simple randomized algorithm that computes witnesses in Õ(nω ) time. In Section 2 we describe a deterministic algorithm for computing the witnesses in Õ(nω ) time. It is esse... |

29 |
Simulating (logn) c -wise Independence in NC
- Berger, Rompel
- 1991
(Show Context)
Citation Context ... since each probability space is small the total work is increased by only a polylogarithmic factor. We note that the two approaches described above had been combined in a different way previously in =-=[18, 7, 22]-=-. The random variables used there were k-wise independent resulting in a probability space of size O(n k ). This probability space was then searched using an estimator function in O(k log n) steps. Th... |

20 | Implicit O(1) probe search
- Fiat, Naor
- 1993
(Show Context)
Citation Context |

17 |
Nonoblivious hashing
- Fiat, Naor, et al.
- 1992
(Show Context)
Citation Context |

16 |
A polynomial time generator for minimal perfect hash functions
- Sager
(Show Context)
Citation Context |

11 |
Witnesses for Boolean matrix multiplication and for transitive closure
- Galil, Margalit
- 1993
(Show Context)
Citation Context ...probabilities. A different, more complicated algorithm for this problem, whose running time is slightly inferior, i.e. not Õ(nω ) (but is also O(nω+o(1) )), has been found by Galil and Margalit [20], =-=[14]-=-. The main motivation for studying the computation of witnesses for Boolean matrix multiplication is the observation of Galil and Margalit that this problem is crucial for the design of efficient algo... |

10 |
Perfect hashing and probability
- Nilli
- 1994
(Show Context)
Citation Context ...g h ∈ H that is perfect for S Perfect hash functions have been investigated extensively (see e.g. [9, 10, 11, 12, 16, 19, 21, 26, 27, 30]). It is known (and not too difficult to show, see [11], [16], =-=[24]-=-) that the minimum possible number of bits required to represent such a mapping is Θ(n + log log m) for all m ≥ 2n. Fredman, Komlós and Szemerédi [12] developed a method for constructing perfect hash ... |

3 |
The effect of table expansion on the program complexity of perfect hash functions
- Mairson
- 1992
(Show Context)
Citation Context |

2 |
A faster algorithm for the all pairs shortest path problem for undirected graphs
- Galil, Margalit
- 1991
(Show Context)
Citation Context ...s with small integer weights which are based on fast matrix multiplication. Efficient algorithms for computing the distances in this way were initiated in [3] and improved (for some special cases) in =-=[13]-=-, [28]. The attempt to extend this method for computing the shortest paths as well leads naturally to the above problem, which already found other (related) applications as well. See [4, 14, 20, 28] f... |