ECF - Evolutionary Computation Framework

ECF is a C++ framework intended for application of any type of evolutionary computation. Current features include:

•  parameterless: genotype (individual structure) is the only mandatory parameter
•  genetic algorithm genotypes (bitstring, binary encoded real values, floating point vectors, permutation vectors), genetic programming genotype (tree)
•  individuals may contain any genotypes in any number
•  algorithms: steady state tournament, generational roulette-wheel, particle swarm optimization (PSO)
•  parallel execution in many models (global paralel EA, distributed EA, hybrid parallel EA...) using MPI
•  configurable environment: changing algorithm, genotypes and parameters without recompilation
•  some parameters: various termination criteria, population size and structure, crx and mutation operators selection and usage rate, migration

More detalied feature list

Download

•  current package (v 1.0, 2010/09/16): ECF_1.0.zip, ECF_1.0.tar.gz
•  the latest development version can be found at svn://smaug.zemris.fer.hr/ecf/trunk
•  latest SVN version can be downloaded and installed with this script (thanks to Jernej Makovsek) tested on Ubuntu

 What's new in SVN:

•  new Binary genotype crossover and mutation operators (rev 343)
•  XCS algorithm implementation (rev 342)

Documentation

These files are also included in the package (./help directory).

•  Install: installation and how to build the examples
•  Tutorial: how to use ECF
•  doxygen produced reference: browse, download (zip)
•  a short FAQ (complement to the Tutorial)

Support: Domagoj Jakobović (domagoj.jakobovic @ fer.hr)

Development in progress

•  TODO list
•  cartesian GP
•  installation with CMake
•  parallelization using MPI
•  new algorithms: genetic annealing, LCS, DE
•  ECF in Java

Component authors

(Alphabetical order): Vinko Bedek, Igor Bespaljko, Iva Brajer, Luka Donđivić, Luka Franov, Zvonimir Fras, Danko Komlen, Ivan Kokan, Maja Legac, Tomislav Novak, Lovro Paić-Antunović, Dražen Popović, Ivana Stokić

Project leader & support contact: Domagoj Jakobović (domagoj.jakobovic @ fer.hr)

Acknowledgements

ECF software uses the following components:

•  some features of boost C++ library
•  a simple but very useful XML parser

Additionally, this project was inspired by OpenBEAGLE, a comprehensive OO evolutionary framework.

Last update: 14.06.2013