Reducing Bloat in GP with Multiple Objectives

摘要: This chapter investigates the use of multiobjective techniques in genetic programming (GP) order to evolve compact programs and reduce effects caused by bloating. The underlying approach considers program size as a second, independent objective besides functionality, several studies have found this concept be successful reducing bloat. Based on one specific algorithm, we demonstrate principle GP show how apply Pareto-based strategies GP. outperforms four classical bloat with regard both convergence speed produced an even-parity problem. Additionally, investigate question why can more effective than alternative test problems. analysis falsifies hypothesis that small but less functional individuals are kept population act building blocks for larger correct solutions. leads conclusion advantages probably due increased diversity population.