期刊名称：Current Journal of Applied Science and Technology
印刷版ISSN：2457-1024
出版年度：2018
卷号：25
期号：2
页码：1-9
语种：English
出版社：Sciencedomain International
摘要：Stochastic Local Search (SLS) algorithms are of great importance to many fields of Computer Sciences and Artificial Intelligence. This is due to their efficient performance when applied for solving randomly generated satisfiability problems (SAT). Our focus in the current work is on one of the SLS dynamic weighting approaches known as multi-level weight distribution (mulLWD). We experimentally investigated the performance and the weight behaviors of mulLWD. Based on our experiments, we observed that the 2nd level weights movements could lead to poor performance of mulLWD, especially when applied for solving large and harder SAT problems. Therefore, we developed a new heuristic that could reduce the cost of the 2nd level neighborhood exploitation known as partial multi-level weight distribution mulLWD+. Experimental results indicate that mulLWD+ heuristic has significantly better performance than mulLWD in a wide range of SAT problems.
