Abstract. In recent years, dynamic local search (DLS) clause weighting algorithms have emerged as the local search state-of-the-art for solving propositional satisfiability problems. This paper introduces a new approach to clause weighting, known as Divide and Distribute Fixed Weights (DDFW), that transfers weights from neighbouring satisfied clauses to unsatisfied clauses in order to break out from local minima. Unlike earlier approaches, DDFW continuously redistributes a fixed quantity of weight between clauses, and so does not require a weight smoothing heuristic to control weight growth. It also exploits inherent problem structure by redistributing weights between neighbouring clauses.To evaluate our ideas, we compared DDFW with two of the best reactive local search algorithms, AdaptNovelty+ and RSAPS. In both these algorithms, a problem sensitive parameter is automatically adjusted during the search, whereas DDFW uses a fixed default parameter. Our empirical results show that DDFW has consistently better performance over a range of SAT benchmark problems. This gives a strong indication that neighbourhood weight redistribution strategies could be the key to a next generation of structure exploiting, parameter-free local search SAT solvers.
Abstract. In recent years, dynamic local search (DLS) clause weighting algorithms have emerged as the local search state-of-the-art for solving propositional satisfiability problems. However, most DLS algorithms require the tuning of domain dependent parameters before their performance becomes competitive. If manual parameter tuning is impractical then various mechanisms have been developed that can automatically adjust a parameter value during the search. To date, the most effective adaptive clause weighting algorithm is RSAPS. However, RSAPS is unable to convincingly outperform the best non-weighting adaptive algorithm AdaptNovelty + , even though manually tuned clause weighting algorithms can routinely outperform the Novelty + heuristic on which AdaptNovelty + is based. In this study we introduce R+DDFW + , an enhanced version of the DDFW clause weighting algorithm developed in 2005, that not only adapts the total amount of weight according to the degree of stagnation in the search, but also incorporates the latest resolution-based preprocessing approach used by the winner of the 2005 SAT competition (R+AdaptNovelty + ). In an empirical study we show R+DDFW + improves on DDFW and outperforms the other leading adaptive (R+Adapt-Novelty + , R+RSAPS) and non-adaptive (R+G 2 WSAT) local search solvers over a range of random and structured benchmark problems.
For decades, the use of weights has proven its superior ability to improve dynamic local search weighting algorithms’ overall performance. This paper proposes a new mechanism where the initial clause’s weights are dynamically allocated based on the problem’s structure. The new mechanism starts by examining each clause in terms of its size and the extent of its link, and its proximity to other clauses. Based on our examination, we categorized the clauses into four categories: (1) clauses small in size and linked with a small neighborhood, (2) clauses small in size and linked with a large neighborhood, (3) clauses large in size and linked with a small neighborhood, and (4) clauses large in size and linked with a large neighborhood. Then, the initial weights are dynamically allocated according to each clause category. To examine the efficacy of the dynamic initial weight assignment, we conducted an extensive study of our new technique on many problems. The study concluded that the dynamic allocation of initial weights contributes significantly to improving the search process’s performance and quality. To further investigate the new mechanism’s effect, we compared the new mechanism with the state-of-the-art algorithms belonging to the same family in terms of using weights, and it was clear that the new mechanism outperformed the state-of-the-art clause weighting algorithms. We also show that the new mechanism could be generalized with minor changes to be utilized within the general-purpose stochastic local search state-of-the-art weighting algorithms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.