The problem representation defines the searching space in Evolutionary Computation (EC). Recent research and developments of EC in optimisation show its robustness and capability of solving different types of optimisation problem due to its stochastic characteristic. As all other computer intelligence system, EC treats an optimisation problem as a searching problem, and the main objective is to find the optimal parameters (solution), or combination of parameters for the problem, and EC simulates the process of Biological Evolution, where each solution evolve/mutate/crossover to generate a better solution.
Traditional Evolutionary Algorithm only involve a single species (single searching space) in optimisation process. In current co-evolutionary (both competitive and cooperative) methodology, a searching space (single searching space) is further divided into sub space and the searching process is carried out in a divide and conquer fashion.
In Co-evolving island models with different species, there are more than one species (more than one searching space) involving in the optimisation process. The alternative searching space can be theoretically derived, such as different colour space (RGB and HSI), mathematically transformed (Fourier and Laplace Transform), dynamically evolve during the searching space, or even as simple as rescaling the space domain.
These alternative spaces are then assisting the objective space in a cooperative co-evolution manner.
Such cooperative process could be as simple as a migration technique or as hard as a host-parasite
(solution from alternative space live on top of solution from objective space) relationship with
entropy test. There are still many design issues that are not investigated in this area.
