On the Query Complexity of Clique Size and Maximum Satisfiability

作者:

Highlights:

摘要

This paper explores the bounded query complexity of approximating the size of the maximum clique in a graph (Clique Size) and the number of simultaneously satisfiable clauses in a 3CNF formula (MaxSat). The results in the paper show that for certain approximation factors, approximating Clique Size and MaxSat are complete for corresponding bounded query classes under metric reductions. The completeness result is important because it shows that queries and approximation are interchangeable: NP queries can be used to solve NP-approximation problems and solutions to NP-approximation problems answer queries to NP oracles. Completeness also shows the existence of approximation preserving reductions from many NP-approximation problems to approximating Clique Size and MaxSat (e.g., from approximating Chromatic Number to approximating Clique Size). Since query complexity is a quantitative complexity measure, these results also provide a framework for comparing the complexities of approximating Clique Size and approximating MaxSat. In addition, this paper examines the query complexity of the minimization version of the satisfiability problem, MinUnsat, and shows that the complexity of approximating MinUnsat is very similar to the complexity of approximating Clique Size. Since MaxSat and MinUnsat share the same solution space, the “approximability” of MaxSat is not due to the intrinsic complexity of satisfiability, but is an artifact of viewing the approximation version of satisfiability as a maximization problem.

论文关键词:

论文评审过程:Received 10 January 1996, Available online 25 May 2002.

论文官网地址:https://doi.org/10.1006/jcss.1996.0070