Using knowledge-based system with hierarchical architecture to guide the search of evolutionary computation

Regional knowledge is determined by function's fitness landscape patterns, such as basins, valleys and multi-modality. Furthermore, for constrained optimization problems, the knowledge of feasible/infeasible regions can also be regards as regional knowledge. Therefore, it would be very helpful if there were a general tool to allow for the representation of regional knowledge, which can be acquired from evolutionary search and then be in reverse applied to guide the search. We define region-based schemata, implemented as belief-cells, which can provide an explicit mechanism to support the acquisition, storage and manipulation of the regional knowledge of a function landscape. In a cultural algorithm framework, the belief space can "contain" a set of these schemata, which can be arranged in a hierarchical architecture, and can be used to guide the search of the evolving population, i.e. region-based schemata can be used to guide the optimization search in a direct way by pruning the infeasible regions and promoting the promising regions. The experiments for an engineering problem with nonlinear constraints indicate the potential behind this approach.