Genetic Programming is an optimizing strategy inspired by genetic and evolution theory.
GPM is a genetic programming library written in C++17. It started as a modernisation of https://github.com/Ambrosys/gpcxx , the goal was to replace object polymorphism with std::variant polymorphism. Aa a result of benchmarking, it became clear that std::variant did not have runtime advantages (in the tested use case). Even std::variant offers value semantic and double dispatch abilities. The decision was made against using std::variant as the base for the tree structure.
- std::variant lacks some features for recursive data structures (the benchmarks used boost::variant as a workaround) https://stackoverflow.com/questions/39454347/using-stdvariant-with-recursion-without-using-boostrecursive-wrapper
- our variant implementation of the tree leaked implementation details to the users The current goal is to implement gpm with nodes in a continues chunk of memory and minimal overhead inspired the way binary heaps are implemented.
- use modern C++ futures to make GP more easily accessible with C++.
Requiements
- cmake
- gcc (tested with gcc-7) or clang (tested with clang-6)
Procedure
git clone https://github.com/gchoinka/gpm.git
mkdir gpm/build && cd gpm/build
cmake .. && cmake --build .
cmake --build . --target run_tree_benchmarkBrowse the documentation.