5.2. Algorithms customization
Among all the previous SI algorithms, the search strategy, named “diversification VS intensification” is the most important core driving the changes of the search process. The term diversification means the exploration of the search space, while the term intensification indicates the exploitation of the accumulated search experience (Consoli and Darby-Dowman, 2007). In other words, diversification and intensification can be understood as the exploration of the whole search space (global search) and the exploitation of certain promising area (local search) respectively. For example, when the search process starts, it generates and computes a set of random individuals in the search domain in order to find the promising areas (diversification). Then the algorithm needs to investigate the promising zones to find the local optimum point (intensification). Nevertheless, the forces of diversification and intensification mutually interact. Even within one search step, some actions may behave as a diversified search, while others intensively search the current area. Finding a good balance between diversification and intensification is essential in order to quickly identify regions in the search space with high quality solutions, without wasting too much time in the region with low quality solutions.
In general, the population size can be treated as a straight-forward sign of diversification, while the number of iterations can be regarded as a symbol of intensification. Since a larger population means a more diversified search and more rounds of iteration suggest a more intensified search in one area. However, it does not means that the combination of larger population and more iteration can certainly lead to better solutions, as larger populations may cause a low convergent speed and more iteration may result into trapping at a local optimum. The trade-off between diversification and intensification is due to the reciprocal of the various intrinsic components, which are named I&D (intensification & diversification) components, as any algorithmic or functional component (operators, actions, strategies, etc.) that has a diversification and/or intensification effect on the search process. Therefore, a unified view named I&D frame on critical inherent components of different algorithms shown in Fig. 8 may be helpful to underline the similarities and differences between them and provide practical insights for the hybridization of diverse algorithms (Blum and Roli, 2003).