Graph representations in genetic programming

Abstract Graph representations promise several desirable properties for genetic programming (GP); multiple-output programs, natural of code reuse and, in many cases, an innate mechanism neutral drift. Each graph GP technique provides a program representation, operators and overarching evolutionary algorithm. This makes it difficult to identify the individual causes empirical differences, both between these methods comparison traditional GP. In this work, we empirically study behaviour Cartesian (CGP), linear (LGP), evolving graphs by By fixing some aspects configurations, performance each method combination with three different EAs: generational, steady-state $$(1+\lambda )$$ ( 1 + λ ) . general, find that best choice operator algorithm depends on problem domain. Further, can increase search complex real-world regression problems particularly ( $$1 + \lambda$$ ) EA, are significantly better digital circuit synthesis tasks. We further show intermediate results tuning LGP’s number registers CGP’s levels back parameter is utmost importance contributes convergence optimization when solving benefit from reuse.