The success of search-based optimisation algorithms depends on appropriately balancing exploration and exploitation mechanisms during the course of the search. We introduce a mechanism that can be used with Differential Evolution (DE) algorithms to adaptively manage the balance between the diversification and intensification phases, depending on current progress. The method—Similarity-based Neighbourhood Search (SNS)—uses information derived from measuring Euclidean distances among solutions in the decision space to adaptively influence the choice of neighbours to be used in creating a new solution. SNS is integrated into explorative and exploitative variants of JADE, one of the most frequently used adaptive DE approaches. Furthermore, SHADE, which is another state-of-the-art adaptive DE variant, is also considered to assess the performance of the novel SNS. A thorough experimental evaluation is conducted using a well-known set of large-scale continuous problems, revealing that incorporating SNS allows the performance of both explorative and exploitative variants of DE to be significantly improved for a wide range of the test-cases considered. The method is also shown to outperform variants of DE that are hybridised with a recently proposed global search procedure, designed to speed up the convergence of that algorithm.
- Differential evolution
- Diversity management
- Global search
- Large-scale continuous optimization