Semantic Genetic Programming

Abstract

Semantic genetic programming (SGP) is a relatively new thread in GP research, which originated in the immense complexity of the genotype phenotype mapping in evolutionary program synthesis. As discussed in Sect. 1.4, minor modifications of program code may result in fundamentally different behavior; on the other hand, an overhaul of program may leave its behavior intact. The relationship between program source code (syntax) and its behavior (semantics for the sake of this chapter) is very complex. SGP germinated from the increasing belief that to make evolutionary program synthesis scalable, program synthesis algorithms need to explicitly take program semantics into account. In this chapter, we provide a concise insight into SGP and show how its conceptual underpinnings relate to behavioral program synthesis and execution records.