Genetic Algorithm for Octave
An expansion of the Genetic Algorithm package for GNU Octave that supports parallelisation and bounds.
This is a fork of the GNU Octave GA package for genetic algorithms, found at https://octave.sourceforge.io/ga/
Here is a version of the package with an expanded feature set that we developed for a project. This package does not perfectly implement the corresponding MATLAB package, but it implements most of the key features.
Added features include:-
- supoort for parallelization, taking advantage of multicore/multithreaded architecture.
- the support of upper and lower bounds on the variables to be optimised.
- support of the 'OutputFcns' feature that enables a user defined function to be applied to the population at each generation.
Dependencies
- The parallelization requires the parallel (https://octave.sourceforge.io/parallel/) package.
Usage
- See https://octave.sourceforge.io/ga/function/ga.html However, not all the features listed in the sourceforge package have been implemented.
gaoptimset
Example
options = gaoptimset('Generations', 80, 'PopulationSize', 16, 'InitialPopulation', IP, 'OutputFcns', myoutputfun, 'UseParallel', 'always', 'Vectorized', 'off');
Parameter-Value pairs
| Parameter | Value | Default |
|---|---|---|
| 'CreationFcn' | UserDefinedFuncton | @gacreationuniform |
| 'CrossoverFcn' | UserDefinedFunction | @crossoverscattered |
| 'CrossoverFraction' | Decimal | 0.8 |
| 'EliteCount' | Integer | 2 |
| 'FitnessLimit' | Double | -Inf |
| 'FitnessScalingFcn' | UserDefinedFunction | @fitscalingrank |
| 'Generations' | Integer | 200 |
| 'InitialPopulation' | Array | [] |
| 'InitialScores' | Array | [] |
| 'MutationFcn' | {UserdefinedFunction, Scale, Shrink} | {@mutationgaussian, 1, 1} |
| 'OutputFcns' | UserDefinedFunction | [] |
| 'PopInitRange' | Array | [-10; 10] |
| 'PopulationSize' | Integer | 20 |
| 'SelectionFcn' | UserDefinedFunction | @selectionstochunif |
| 'TimeLimit' | Double | Inf |
| 'UseParallel' | 'always' or 'never' | 'never' |
| 'Vectorized' | 'on' or 'off | 'off' |
ga
[x, fval, exitflag, output, population, scores] = ga (fitnessfcn, nvars, A, b, Aeq, beq, LB, UB, nonlcon, options)
Implemented
- fitnessfcn.
- nvars.
- LB.
- UB.
- options.
Not implemented
- exitflag
- The inequality A*x<=b.
- The equality Aeq*x=beq.
- The nonlinear constraints, nonlcon.