Sugal

Sugal [soo-gall] is the Sunderland Genetic Algorithm System. It is the most powerful free package for this purpose. It can be used for problems that are difficult to solve with traditional optimization techniques, including problems that are not well defined or are difficult to model mathematically.

Major features of the Sugal

  • Initialisation: Uniform, Gaussian, Loaded Bit and from File.
  • Selection: Roulette, Integral Roulette (i.e. Expected Value Model), Tournament and Uniform .
  • Fitness Normalisation: Biased, Inversion, Linear and Mean-Linear, Rank Linear and Rank Geometric fitness normalisation (all either for minimisation or maximisation of fitness function).
  • Crossover: Uniform, One-Point, Two-Point and arbitrary N-Point crossover, variable crossover rates, multiple operator application.
  • Mutation: Bit Inversion, Gene Reinitialisation, Step-delta, Uniform-delta and Gaussian-delta Mutation, including time-decay mutation sizes; variable mutation rates, multiple operator application.
  • Replacement: Uniform, Rank-based, Tournament, Crowding and Parental replacement; replacement can be Unconditional, Simple Conditional (i.e. If Improved) or Simulated Annealing; variable replacement rate (i.e. generation gap); elitism.
  • Stopping conditions: number of generations, fitness target achieved, fitness or diversity convergence level.

The Sugal Genetic Algorithm subsumes most of the GA models known by the end of 1990-th. This is very powerful package, but it also have expanded requirements to user (e.g. knowledge of C++).

Product information

  • Current Version: 2.1
  • License Type: Free for Personal and Non-Profit Use
  • Targeted Platforms: Tested on Sun Solaris 2 and PC Borland C++ 3.x, and has been successfully run on Windows, OS/2 and SGI. Sugal has platform-independent Graphical User Interface