Cartesian Genetic Programming in Julia. Based on the Cambrian.jl framework.
CartesianGeneticProgramming.jl can be installed through the Julia package manager
pkg> add CartesianGeneticProgramming.jlExamples of usage are provided in scripts. To run any of these examples, do
$ julia scripts/gym.jlThis will create and populate gens and logs folders, which store CGP
individuals and evolutionary logs, respectively.
Note that these examples have only been tested on Linux and have additional
dependencies which must be installed, like PyCall or
ArcadeLearningEnvironment.
CGP.jl comes with tests which can offer examples of detailed usage of the
different genetic operators and CGP extensions. To run all tests, use
pkg> test CartesianGeneticProgrammingConfiguration is handled by src/config.jl. Configurable options are read in
through YAML files (an example being cfg/test.yaml). All possible node
functions are defined in src/functions.jl and are populated in the
configuration files.
This repository is now intended only for floating point inputs and therefore no longer implements Mixed-Type CGP. The commit reference for the results in [1] is 096e843, however this requires Julia 0.6 and may no longer function. MTCGP is currently being rewritten at https://github.com/d9w/MTCGP.jl. This is done to separate heavy image processing package dependencies which are not necessary for base CGP.
[1] Wilson, Dennis G., et al. "Evolving simple programs for playing Atari games." Proceedings of the Genetic and Evolutionary Computation Conference. ACM, 2018.