Article Preview
TopIntroduction
During the last decades, there has been a huge growing in the area of evolutionary computing, which has indicated modern techniques for solving various types of optimization problems. In contrast to classical or traditional optimization approaches, which emphasize exact and accurate computation, but these approaches may fall down in obtaining the solution of a global optimum. Based on that, evolutionary computation emerged to provide more efficient and robust technique for solving complex problems (Shehadeh et al. 2018a; Fogel, 2005). “Genetic algorithm (GA)” is one of the most prevalent branches between the existing evolutionary approaches. GA is considered as an inherently discrete stochastic search approach depends on the techniques of generating genetics, natural selection, and evolution (Shehadeh et al, 2018a; Paulinas and Ušinskas, 2007). Since this type of metaheuristics simultaneously can search and evaluate many sites in the domain of problem, which is more likely to discover the global optimum of an optimization problem. Furthermore, it easy to implement and use, which uses a measure of performance of simple scalar that does not use or require derivative information (Shehadeh et al., 2018a).
More recently, Hisham A. Shehadeh et al. (Shehadeh et al., 2017; Shehadeh et al. 2018b, Shehadeh et al. 2018c) proposed a new metaheuristic method, namely ‘‘Sperm Swarm Optimization’’ (denoted as SSO). The idea of this approach is inspired by the behaviors of sperm swarm through the procedure of natural fertilization. The theory of SSO represents a solution process, which each sperm swims through the domain of multidimensional search space while the sperm’s position and velocity are constantly updated based on the previous position of the sperm, as well as the best performance of the swarm in the entire population.
In contrast with GA, SSO is an inherently continuous approach in which has various attractive features. It has memory, so the prior knowledge can be retained by all the sperms in each generation; whereas in GA, the scenario is different, which is considered as an inherently discrete approach, so the prior knowledge of the problem is discarded each iteration by reserving the best individuals and eliminating the worst individuals at each generation. To date, SSO has been successfully applied to generate the optimal solution for different continuous nonlinear functions in practice (Shehadeh et al., 2018c), but until recently it had not been hybridized to deal with multimodal problems. SSO seems particularly appropriate for multimodal tasks mainly because of the good quality of solutions and the high speed of convergence that the algorithm shows for solving different kinds of single-objective problems (Shehadeh et al., 2018c).
There are many studies that have been investigated the hybridization of evolutionary algorithms with local search. We can summarize them as follows:
Soleimani et al. proposed a hybrid approach that integrates “Particle Swarm Optimization (PSO)” with GA. The proposed approach was used to optimize problems of supply chain network. The results showed that the proposed approach has better convergence and quality of solution than GA (Soleimani & Kannan, 2015).
On the other hand, Samuel et al. suggested new optimization approach that merges the functionality of PSO and GA. The proposed approach was used to solve the scheduling problem of power generator. The results proved the efficiency of the proposed approach in finding solution for the aforementioned problem (Samuel & Rajan, 2015).
In different view, Fang et al. proposed a new hybrid approach that integrates “Artificial Fish Swarm Algorithm (AFSA)” with GA. This approach was used to solve scheduling problem of the hydrothermal systems. To prove the performance of the proposed approach, Fang et al. tested this approach on two hydrothermal systems (Fang et al., 2014).
In a different work, Kao et al. proposed a hybrid approach that integrates PSO with GA. The proposed approach was applied to solve 17 multi-modal functions. The results showed that the proposed approach has better convergence and quality of solution than other approaches (Kao & Zahara, 2008).
Gholami et al. discussed a hybrid approach that merges the functionality of PSO with GA. They used this approach to optimize bank shape problem. The result showed the performance of the approach under different scenarios (Gholami et al., 2018).