GPFramework (0.1)

An extensible Java framework for tree-based Genetic Programming

Introduction

GPFramework is a flexible Java framework for quick prototyping of Genetic Programming systems. Its main purpose has been, so far, to study Genetic Programming computational complexity from an experimental point of view, hence it is by no means a complete Genetic Programming framework, but might evolve to become that.

Currently GPFramework handles mutation-only Genetic Programming.

How to use it

As a result of our work in the analysis of Genetic Programming's complexity, GPFramework includes a number of fitness functions for sorting, order and majority. It currently implements two underlying mechanisms to evolve programs:

• SMO-GP, an algorithm inspired by the SEMO evolutionary multi-objective algorithm,
• (m+m)-GP, a generic population-based evolutionary algorithm where the size of the

  population can be set as a parameter

with several selection criteria including:

• Parsimony pressure,
• strict/weak hill climbing,
• multi-objective (for SMOGP).

The best way to learn how to use it is to read the wide inline documentation (javadoc) and have a look to how sorting, order and majority have been implemented.

GPFramework is distributed together with a NetBeans project, but should be opened fairly easily in other IDEs as well.

Dependencies

Three libraries are needed in order to make GPFramework compile.

• Colt, http://acs.lbl.gov/software/colt
• Apache Commons Math >= 1.2
• Apache Commons CLI >= 2.2

It is sufficient to add those to the project in order to use GPFramework.

Licensing

This software is licensed under the MIT License. This means that you can use it for whatever purpose, without any licensing limitation on the derivatives. See LICENSE file for more details.

Tommaso Urli and Markus Wagner