Authors: | Matijević, Luka | Affiliations: | Mathematical Institute of the Serbian Academy of Sciences and Arts | Title: | General Variable Neighbourhood Search Approach to MAX-3SAT Problem | First page: | 29 | Last page: | 30 | Conference: | The 8TH International Conference on Logic and Applications - LAP 2019 | Issue Date: | 2019 | Rank: | M34 | URL: | http://imft.ftn.uns.ac.rs/math/cms/uploads/Main/LAP_2019_Book_of_Abstracts.pdf | Abstract: | MAX-3SAT problem is a version of MAX-SAT problem where every clause has exactly three literals. It is one of the most important problems of compu- tational complexity theory, and as such, many solvers have been developed for it. Since it belongs to the class of NP-complete problems, it is usually solved by implementing heuristic methods. In this paper we used general variable neighbourhood search (GVNS) to obtain the best possible solution in a given amount of time. GVNS is neighbor- hood based search algorithm that involves two main steps: perturbation and improvement. We defined two sets of neighbourhoods, each of them used in the perturbation step with a predefined probability p. For the improvement step we used variable neighbourhood descent (VND), which is a local search heuristic that explores several neighborhood structures in a deterministic way. The proposed GVNS approach is tested on a set of benchmark instances found at: https://www.cs.ubc.ca/ hoos/SATLIB/benchm.html, and compared with state-of-the-art WalkSAT implementation: http://www.cs.rochester.edu/u/kautz/walksat/. We concluded that for smaller instances our approach gives similar results as aforementioned WalkSAT imple- mentation, but it performed better for larger instances, given that it finds very good solutions very quickly. |
Keywords: | satisfiability problem | NP-completeness | incomplete solvers | stochastic algorithms | metaheuristic approach | Publisher: | University Center Dubrovnik, Croatia, | Project: | Representations of logical structures and formal languages and their application in computing |
Show full item record
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.