Abstract
One of the major problems algorithm designers usually face is to know in advance whether a proposed optimization algorithm is going to behave as planned, and if not, what changes are to be made to the way new solutions are examined so that the algorithm performs nicely. In this work we develop a methodology for differentiating good neighborhoods from bad ones. As a case study we consider the structure of the space of assignments for random 3-SAT formulas and we compare two neighborhoods, a simple and a more refined one that we already know the corresponding algorithm behaves extremely well. We give evidence that it is possible to tell in advance what neighborhood structure will give rise to a good search algorithm and we show how our methodology could have been used to discover some recent results on the structure of the SAT space of solutions. We use as a tool “Go with the winners”, an optimization heuristic that uses many particles that independently search the space of all possible solutions. By gathering statistics, we compare the combinatorial characteristics of the different neighborhoods and we show that there are certain features that make a neighborhood better than another, thus giving rise to good search algorithms.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Carson, T.: Empirical and Analytic Approaches to understanding Local Search Heuristics. PhD thesis, University of California, San Diego (2001)
Carson, T., Impagliazzo, R.: Hill climbing finds random planted bisections. In: Proc. 12th Annual ACM Symposium on Discrete Algorithms, SODA (2001)
Dimitriou, T., Impagliazzo, R.: Towards an analysis of local optimization algorithms. In: Proc. 28th Annual ACM Symposium on Theory of Computing, STOC (1996)
Dimitriou, T., Impagliazzo, R.: Go with the Winners for Graph Bisection. In: Proc. 9th Annual ACM Symposium on Discrete Algorithms (SODA), pp. 510– 520 (1998)
Frank, J.: Learning weights for GSAT. In: Proc. IJCAI 1997, pp. 384–391 (1997)
Garey, M.R., Johnson, D.S.: Computers and Intractability. Freeman, San Francisco (1979)
Gent, I., Walsh, T.: An empirical analysis of search in GSAT. Journal of Artificial Intelligence Research 1, 23–57 (1993)
Gent, I., Walsh, T.: The SAT Phase Transition. In: Proc. of ECAI 1994, pp. 105–109 (1994)
Hoos, H.: On the run-time behavior of stochastic local search algorithms for SAT. In: Proc. AAAI 1999, pp. 661–666 (1999)
Johnson, D.S.: Report on Challenges for Theoretical Computer Science. In: Workshop on Challenges for Theoretical Computer Science, Portland (2000), http://www.research.att.com/~dsj/
Mitchell, D., Selman, B., Levesque, H.J.: Hard and easy distributions of SAT problems. In: Proc. AAAI 1992, San Jose, CA, pp. 459–465 (1992)
Selman, B., Kautz, H.A., Cohen, B.: Local search strategies for satisfiability testing. In: Second DIMACS Challenge on Cliques, Coloring and Satisfiability (October 1993)
Selman, B., Levesque, H.J., Mitchell, D.: A new method for solving hard statisfiability problems. In: Proc. AAAI 1992, San Jose, CA (1992)
Schuurmans, D., Southey, F.: Local search characteristics of incomplete SAT procedures. Artif. Intelligence 132(2), 121–150 (2001)
Wu, Z., Wah, W.: Trap escaping strategies in discrete Langrangian methods for solving hard satisfiability and maximum satisfiability problems. In: Proc. AAAI 1999, pp. 673–678 (1999)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2004 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Dimitriou, T., Spirakis, P. (2004). How to Tell a Good Neighborhood from a Bad One: Satisfiability of Boolean Formulas. In: Ribeiro, C.C., Martins, S.L. (eds) Experimental and Efficient Algorithms. WEA 2004. Lecture Notes in Computer Science, vol 3059. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-24838-5_15
Download citation
DOI: https://doi.org/10.1007/978-3-540-24838-5_15
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-22067-1
Online ISBN: 978-3-540-24838-5
eBook Packages: Springer Book Archive