Likelihood-Based Random Search Technique for Solving Unconstrained Optimization Problems

摘要

Although many methods have been devised for solving optimization problems, there still a pressing need for more effective and efficient techniques. Most of the proposed techniques are effective in solving the optimization problems. They, however, fall short when dealing with specific problems (e.g. problems with multiple local optima). This paper offers an innovative technique for optimization problems. The proposed method combines between the random-guided search and both techniques for identifying the promising regions of the search space and mapping techniques that bias the search to these promising regions; thereby quickly finding the global minimum values. Experiments with our prototype implementation showed that our method is effective in finding exact or very close approximation of the global minimum values for challenging functions obtained from well-known benchmarks. Our comparative study showed that our method is superior to other state-of-art methods.