Genetic Programming with Local Hill-Climbing

This paper proposes a new approach to Genetic Programming (GP). In traditional GP, recombination can cause frequent disruption of building-blocks or mutation can cause abrupt changes in the semantics. To overcome these difficulties, we supplement traditional GP with a recovery mechanism of disrupted building-blocks. More precisely, we integrate the structural search of traditional GP with a local hill-climbing search, using a relabeling procedure. This integration allows us to extend GP for Boolean and numerical problems. We demonstrate the superior effectiveness of our approach with experiments in Boolean concept formation and symbolic regression.