Tree-based Genetic Programming

Genetic programming is part of the evolutionary computation subfield of artificial intelligence. Genetic programming uses techniques inspired by biological evolution to evolve computer programs that perform user-defined tasks. The aim of OakGP is to provide an efficient and flexible open-source framework to allow people to experiment with genetic programming. To support this aim OakGP incorporates the following features:

• Tree-based. Programs are represented using tree data structures.
• Type information. Functions, variables and constants all have associated type information.
  • Checked during the construction of tree structures to ensure type safety.
  • Allows user-defined types.
  • Supports generic types.
• Pure functions. OakGP provides a wide range of functions that can be used in the evaluation of tree structures. Function domains include:
  • Arithmetic - add, subtract, multiply and divide functions for int, long and double primitives, plus java.math.BigDecimal and java.math.BigInteger.
  • Higher order functions - map, filter, reduce.
  • Comparison operators.
• Genetic operators. OakGP uses techniques inspired by biological evolution to evolve new tree structures. The techniques include two main sub-categories:
  • Mutation - hoist, point, shrink and subtree mutation.
  • Crossover - common region and subtree crossover.
• Configurable. Allows configuration of the following:
  • Primitive set (functions, constants and variables).
  • Termination criteria.
  • Functions.
  • Genetic operators.
  • Selection strategies (e.g. fitness proportionate selection or rank selection).
• Extensible. Framework enables extension (i.e. adding new functions, genetic operators, etc.) without the need to modify the existing code. The open-source nature of the project, combined with the documentation on this site, provides developers with enough information to add new functionality specific to the problems they are trying to solve.
• Documented. This site includes an introduction to genetic programming and a guide to getting started with OakGP.
• Examples. This site provides examples of how OakGP can be used to find solutions for:
  • Symbolic Regression
  • Grid War
  • Towers of Hanoi
  • The Artificial Ant Problem (also known as the Santa Fe Trail problem)
• JVM-based. Implemented in Java, OakGP runs on the Java Virtual Machine.
  • An established and stable platform.
  • Runs on many operating systems.
  • Supported by a large collection of libraries and tools.
• Open-source.
  • Apache License, Version 2.0.
  • Source code is viewable on GitHub.