GEATbx.com

Features

What is the GEATbx ?

• powerful optimization tool using evolutionary algorithms (genetic algorithms, evolution strategies)
• applicable to a broad range of problems and systems
• comprehensive implementation of evolutionary algorithms for Matlab, compare on your own using my overview of Evolutionary Algorithms in Matlab

How does the GEATbx work ?

• implements a wide range of evolutionary operators and principles
• all operators and features are fully integrated into one environment
• runs on every Matlab-supported platform (requires Matlab 5.3 or a more recent version, already tested/used under Matlab 5.3, 6.0, 6.1, 6.5.1, 7.0, 7.1), no platform dependence (runs on Windows, Linux, Solaris, ... without any changes)

Who should use the GEATbx ?

• engineers solving real-world problems
• researchers comparing and developing new algorithms and test functions
• students becoming acquainted with evolutionary algorithms

GEATbx Features

• high level functions to all operators
• broad class of operators
  • selection: linear/non-linear ranking, stochastic universal sampling, truncation, tournament, roulette wheel, local selection
  • recombination: discrete, intermediate, line, extended line, single/double point, shuffle, uniform crossover, reduced surrogate
  • mutation: binary, real valued
• complete support of multi-objective optimization
  • PARETO-ranking and goals attainment (using different implementations for extra speed), see ranking.m
  • sharing (fitness sharing in search space) integrated into ranking
  • collection of non-dominated solutions during the optimization (archive)
  • return of non-dominated solutions in return population
  • visualization of multi-objective individuals and solutions see plotmop.m
  • many multi-objective example functions
• population models: global model, regional model (multiple subpopulations) and local model (local selection and reinsertion, different neighborhood structures)
• migration (regional model): unrestricted, ring, neighborhood
• reinsertion: global, regional and local
• multiple strategy support
• real, integer and binary (linear and logarithmic scaling, gray coding) variable representation
• sophisticated visualization
• comfortable monitoring and storing of results
• incorporation of problem specific knowledge (special initialization and problem specific visualization)
• The GEATbx can be completely compiled into C/C++ Code using the Matlab Compiler.

GEATbx Implementation

• m-file implementation
• compatible on all computer platforms
• modular, user-friendly structure
• high entry functions
• many examples and test functions included (ready to run)
  • see Example Functions Description
  • TSP: travelling sales person (incl. demo function demotsplib.m and specific visualization)
  • VRP: vehicle routing problem (incl. demo function demovrp.m and specific visualization)
• extensive documentation
  • tutorial
  • quick start
  • extra chapter about How to Approach new Optimization Problems in tutorial
  • overview of structure
  • explanation of algorithms
  • full description of different strategies and competing subpopulations
  • description of Multi-objective Ranking (inkl. PARETO, Goal attainment, sharing (and weighted sum))
  • extensive discussion of Combination of Operators and Options to Produce Evolutionary Algorithms in Introduction
  • complete Example Functions Description
  • index to all operators and functions
  • available in HTML and PDF
• default parameter settings
• easily extensible (definition of new operators)
• easy incorporation of problem-specific functions for initialization and visualization
• fully compatible with Matlab 5.3, 6.0, 6.1, 6.5, 7.1 (and probably newer versions)