Feasibility Structure Modeling: An Effective Chaperone for Constrained Memetic Algorithms

摘要: An important issue in designing memetic algorithms (MAs) is the choice of solutions population for local refinements, which becomes particularly crucial when solving computationally expensive problems. With single evaluation objective/constraint functions necessitating tremendous computational power and time, it highly desirable to be able focus search efforts on regions where global optimum potentially located so as not waste too many function evaluations. For constrained optimization, must either at trough some feasible basin or particular point along feasibility boundary. Presented this paper an instance optinformatics a new concept modeling structure inequality-constrained optimization problems-dubbed modeling-is proposed perform geometrical predictions locations candidate solution space: deep inside any infeasible region, nearby boundary, region. This knowledge may unknown prior executing MA but can mined progresses. As more are generated subsequently stored database, thus approximated accurately. integral part, paradigm incorporating classification-rather than regression-into framework MAs introduced, allowing estimate boundary such that effective assessments whether should experience refinements made. eventually helps preventing unnecessary consequently reducing number evaluations required reach optimum.