Whale Optimization Algorithm Based on Lamarckian Learning for Global Optimization Problems

Abstract

Whale optimization algorithm (WOA) is a population-based meta-heuristic imitating the hunting behavior of humpback whales, which has been successfully applied to solve many real-world problems. Although WOA has a good convergence rate, it cannot achieve good results in finding the global optimal solution of high-dimensional complex optimization problems. The learning mechanism of Lamarckian evolutionism has the advantages of speeding up and strengthening local search. Through this learning mechanism, solutions with certain conditions can acquire higher adaptability with a higher probability by active learning. To enhance the global convergence speed and get better performance, this paper presents a WOA based on Lamarckian learning (WOALam) for solving high-dimensional function optimization problems. First, the population is initialized by good point set theory so that individuals can be evenly distributed in the solution space. Second, the upper confidence bound algorithm is used to calculate the development potential of the individual. Finally, based on the evolutionary theory of Lamarck, individuals with more development potentials are selected to perform the local enhanced search to improve the performance of the algorithm. The WOALam was compared with six variants of WOA on 44 benchmark functions. The experiments proved that the proposed algorithm can balance the global exploring ability and the exploiting ability well. It could obtain better results with fewer iterations and had good convergence speed and accuracy.