RASLite: Enhancing (W)PMS Solvers Through Dynamic Initial Weight Approach

Menghua Jiang,Yin Chen

DOI: https://doi.org/10.1109/cscwd61410.2024.10580189

2024-01-01

Abstract:Maximum Satisfiability (MaxSAT) is a prototypical constrained optimization problem, featuring two generalized forms: Partial MaxSAT (PMS) and Weighted Partial MaxSAT (WPMS), dealing with hard and soft clauses. Stochastic local search (SLS) algorithms prove to be vital for solving (W)PMS. In this work, we observe that traditional (W)PMS local search solvers typically update soft and hard clause weights under the same conditions and maintain consistent initial weights for soft clauses across the local search process. These methods might produce suboptimal solutions and limit the accuracy of the algorithm during the initial search phase. To handle these issues, we propose two ideas. Firstly, we propose a new clause weighting scheme that imposes constraints on modifications to soft clause weights. Secondly, we propose a dynamic weight initialization scheme to dynamically modify initial soft clause weights. Based on these two ideas, we develop a new SLS solver for (W)PMS named RASLite 1 . Extensive experiments show that RASLite significantly outperforms the state-of-the-art (W)PMS SLS solver SATLike 3.0.