Derandomization That Is Rarely Wrong from Short Advice That Is Typically Good
- 729 Downloads
For every ∈ > 0, we present a deterministic log-space algorithm that correctly decides undirected graph connectivity on all but at most 2 n ∈ of the n- vertex graphs. The same holds for every problem in Symmetric Log-space (i.e., SL.
Using a plausible complexity assumption (i.e., that P cannot be approximated by SIZE(p)SAT, for every polynomial p) we show that, for every ∈ > 0, each problem in BPP has a deterministic polynomial-time algorithm that errs on at most 2 n ε of the n-bit long inputs. (The complexity assumption that we use is not known to imply 209-04.
All results are obtained as special cases of a general methodology that explores which probabilistic algorithms can be derandomized by generating their coin tosses deterministically from the input itself. We show that this is possible (for all but extremely few inputs) for algorithms which take advice (in the usual Karp-Lipton sense), in which the advice string is short, and most choices of the advice string are good for the algorithm.
To get the applications above and others, we show that algorithms with short and typically-good advice strings do exist, unconditionally for SL, and under reasonable assumptions for BPP and AM.
KeywordsDeterministic Algorithm Pseudorandom Generator Input Length Universal Sequence Undirected Graph Connectivity
Unable to display preview. Download preview PDF.
- 1.L. Adleman. Two theorems on random polynomial time. In 10th FOCS, pages 75–83, 1978.Google Scholar
- 2.C. Alvarez and R. Greenlaw. A compendium of problems complete for symmetric logarithmicspace. ECCC report TR96-039, 1996.Google Scholar
- 3.R. Aleliunas, R.M. Karp, R.J. Lipton, L. Lovász and C. Rackoff. Random walks, universaltraversal sequences, and the complexity of maze problems. In 20th FOCS, pages 218–223, 1979.Google Scholar
- 4.R. Armoni, M. Saks, A. Wigderson and S. Zhou. Discrepancy sets and pseudorandom generatorsfor combinatorial rectangles. In 37th FOCS, pages 412–421, 1996.Google Scholar
- 5.V. Arvind and J. öbler. On pseudorandomness and resource-bounded measure. In 17thFSTTCS, Springer-Verlag, LNCS 1346, pages 235–249, 1997.Google Scholar
- 6.L. Babai. Trading Group Theory for Randomness. In 17th STOC, pages 421–429, 1985.Google Scholar
- 8.O. Goldreich, N. Nisan and A. Wigderson. On Yao’s XOR-Lemma. ECCC, TR95-050, 1995.Google Scholar
- 9.O. Goldreich, D. Ron and M. Sudan. Chinese Remaindering with Errors. TR98-062, availablefrom ECCC, at http://www.eccc.uni-trier.de/eccc/, 1998.
- 10.O. Goldreich and A. Wigderson. Derandomization that is rarely wrong from short advice thatis typically good. TR02-039, available from ECCC, 2002.Google Scholar
- 12.R. Impagliazzo, V. Kabanets and A. Wigderson. In search of an easy witness: Exponentialversus probabilistic time. In proceedings of 16th CCC, pages 2–12, 2001.Google Scholar
- 13.R. Impagliazzo and A. Wigderson. P=BPP if E requires exponential circuits: Derandomizingthe XOR Lemma. In 29th STOC, pages 220–229, 1997.Google Scholar
- 14.R. Impagliazzo and A. Wigderson. Randomness vs. Time: De-randomization under a uniformassumption. In 39th FOCS, pages 734–743, 1998.Google Scholar
- 16.R.M. Karp and R.J. Lipton. Some connections between nonuniform and uniform complexityclasses. In 12th STOC, pages 302–309, 1980.Google Scholar
- 17.A. Klivans and D. van Melkebeek. Graph Nonisomorphism has Subexponential Size ProofsUnless the Polynomial-Time Hierarchy Collapses. In 31st STOC, pages 659–667, 1998. Toappear in SICOMP.Google Scholar
- 19.N. Nisan, E. Szemeredi, and A. Wigderson. Undirected connectivity in O(log 1.5 n) space. In33rd FOCS, pages 24–29, 1992.Google Scholar
- 22.R. Shaltiel. A log-space extractor for high values of min-entropy. Personal communication, June 2002.Google Scholar
- 25.A. Ta-Shma. Almost Optimal Dispersers. In 30th STOC, pages 196–202, 1998.Google Scholar
- 26.L. Trevisan. Constructions of Near-Optimal Extractors Using Pseudo-Random Generators.In 31st STOC, pages 141–148, 1998.Google Scholar
- 27.A.C. Yao. Theory and Application of Trapdoor Functions. In23rdFOCS, pages 80–91, 1982.Google Scholar
- 28.D. Zuckerman. Randomness-Optimal Sampling, Extractors, and Constructive Leader Election.In 28th STOC, pages 286–295, 1996.Google Scholar