Background

Cartesian Genetic Programming (CGP) is a tree based representation of a computer program introduced by Julian F. Miller and Peter Thomson in 1997. Several open source codes exist that implement some form of CGP, Miller himself maintain a resource page here.

The dcgpy library here introduced, differs in several aspects from all existing implementations: it allows to perform any-order derivatives on the represented programs (hence implementing a differentiable form of genetic programming), it allows to use Python to use runtime scripting, it is thread-safe and it allows to define kernels as generic functors.

• Cartesian Genetic Programming (CGP)
• Differentiable Cartesian Genetic Programming (dCGP)