Can anyone help me for knowing the options of genetic algorithm?

How can i know what the value that i put in genetic algorithm's options ?

[Voss]

https://www.mathworks.com/help/gads/ga.html#mw_4a8bfdb9-7c4c-4302-8f47-d260b7a43e26

Answer 1

[Star Strider]

To list them all —

opts = optimoptions(@ga)

opts = 
  ga options:

   Set properties:
     No options set.

   Default properties:
             ConstraintTolerance: 1.0000e-03
                     CreationFcn: []
                    CrossoverFcn: []
               CrossoverFraction: 0.8000
                         Display: 'final'
                      EliteCount: '0.05*PopulationSize'
                    FitnessLimit: -Inf
               FitnessScalingFcn: @fitscalingrank
               FunctionTolerance: 1.0000e-06
                       HybridFcn: []
         InitialPopulationMatrix: []
          InitialPopulationRange: []
             InitialScoresMatrix: []
                  MaxGenerations: '100*numberOfVariables'
             MaxStallGenerations: 50
                    MaxStallTime: Inf
                         MaxTime: Inf
                     MutationFcn: []
    NonlinearConstraintAlgorithm: 'auglag'
                       OutputFcn: []
                         PlotFcn: []
                  PopulationSize: '50 when numberOfVariables <= 5, else 200'
                  PopulationType: 'doubleVector'
                    SelectionFcn: []
                     UseParallel: 0
                   UseVectorized: 0

Change the ones you need to change. This is a structure, so to display 'FunctionTolerance' refer to it as —

FcnTol = opts.FunctionTolerance

FcnTol = 1.0000e-06

To change it, use a similar approach —

opts.FunctionTolerance = 1E-5

opts = 
  ga options:

   Set properties:
               FunctionTolerance: 1.0000e-05

   Default properties:
             ConstraintTolerance: 1.0000e-03
                     CreationFcn: []
                    CrossoverFcn: []
               CrossoverFraction: 0.8000
                         Display: 'final'
                      EliteCount: '0.05*PopulationSize'
                    FitnessLimit: -Inf
               FitnessScalingFcn: @fitscalingrank
                       HybridFcn: []
         InitialPopulationMatrix: []
          InitialPopulationRange: []
             InitialScoresMatrix: []
                  MaxGenerations: '100*numberOfVariables'
             MaxStallGenerations: 50
                    MaxStallTime: Inf
                         MaxTime: Inf
                     MutationFcn: []
    NonlinearConstraintAlgorithm: 'auglag'
                       OutputFcn: []
                         PlotFcn: []
                  PopulationSize: '50 when numberOfVariables <= 5, else 200'
                  PopulationType: 'doubleVector'
                    SelectionFcn: []
                     UseParallel: 0
                   UseVectorized: 0

.

[noura]

What are they determined based on and how do I know if I need to change them or not?

[Star Strider]

I would leave them as their set default values initially. Only change them if that would improve the results of your optimisation.

[noura]

Sorry for asking again but i'm beginner in matlab How can i test the optimization results?

[Star Strider]

No wories!

Run your ga code and see if the resulkts are essentially what you expect. If they are not, first check your fitness function (or your objective function, depending no how your fitness function is constructed) to be certain it is doing the correct calculations. (One way to do that would be to give it a set of appropriate parameters and then run it to see what it produces. It does not have to be exact — that is the purpose of running the optimisation — however it has to make sense and produce a reasonable result.) After that, run the ga optimisation a few times to see what it cones up with. If some of the results are acceptable although imprecise, set the appropriate options to decrease the tolerances or increase the number of generations and see if you can improve the result. (I have sets of options that I use for specific problems, so when you find a set that works, save it somewhere to use with similar problems.)

[Walter Roberson]

Note: to list all of the options,

orderfields(struct(optimoptions(@ga)))

Warning: Calling STRUCT on an object prevents the object from hiding its implementation details and should thus be avoided. Use DISP or DISPLAY to see the visible public details of an object. See 'help struct' for more information.

ans = struct with fields:
             ConstraintTolerance: 1.0000e-03
                     CreationFcn: []
                    CrossoverFcn: []
               CrossoverFraction: 0.8000
                         Display: 'final'
                      EliteCount: '0.05*PopulationSize'
                    FitnessLimit: -Inf
               FitnessScalingFcn: @fitscalingrank
                FromOptimProblem: 0
               FunctionTolerance: 1.0000e-06
                       GaVersion: 2
                 GacommonVersion: 2
                     Generations: '100*numberOfVariables'
                       HybridFcn: []
                  InitialPenalty: 10
               InitialPopulation: []
         InitialPopulationMatrix: []
          InitialPopulationRange: []
                   InitialScores: []
             InitialScoresMatrix: []
                IntegerTolerance: 1.0000e-05
                  MaxGenerations: '100*numberOfVariables'
             MaxStallGenerations: 50
                    MaxStallTime: Inf
                         MaxTime: Inf
              MigrationDirection: 'forward'
               MigrationFraction: 0.2000
               MigrationInterval: 20
                     MutationFcn: []
              NonlinConAlgorithm: 'auglag'
    NonlinearConstraintAlgorithm: 'auglag'
                    OptionsStore: [1x1 struct]
                       OutputFcn: []
                      OutputFcns: []
                   PenaltyFactor: 100
                         PlotFcn: []
                        PlotFcns: []
                    PlotInterval: 1
                    PopInitRange: []
                  PopulationSize: '50 when numberOfVariables <= 5, else 200'
                  PopulationType: 'doubleVector'
               ProblemdefOptions: [1x1 struct]
                PropertyMetaInfo: [1x1 struct]
                    SelectionFcn: []
                      SolverName: 'ga'
                   StallGenLimit: 50
                       StallTest: 'averageChange'
                  StallTimeLimit: Inf
                       TimeLimit: Inf
                          TolCon: 1.0000e-03
                          TolFun: 1.0000e-06
                     UseParallel: 0
                   UseVectorized: 0
                      Vectorized: 'off'
                         Version: []

[noura]

What about the time it took more than 8 hours and doesn't complete?

[Star Strider]

@noura — That depends on what the problem is, how many parameters are in the model, and the nature of the fitness function itself.

In my optimoptions structure, I include:

'PlotFcn',@gaplotbestf, 'PlotInterval',1

This tells me how the fitness is progressing.

I would have to see (and understand) your fitness function in order to provide anything further.

Answer 2

[Steven Lord]

Since you've said you're a beginner with MATLAB (and presumably Global Optimization Toolbox), I recommend starting off with the tutorials for how to get started with Global Optimization Toolbox on this documentation page. There are also three examples linked from that page (click the Examples link just below the blue Help Center bar running across the screen, direct link is here.) The examples don't go into the various options available to the various solvers in that much depth, as they're intended to help you get started and at least at first you may not need to change those options.

There are a few examples in the Genetic Algorithm category that go into a little more depth about those options. I'm looking specifically at "Effects of Genetic Algorithm Options". To get to that category, click on "Genetic Algorithm" on the "Get Started" Examples page (the second link above, direct link to those examples here.)

Finally, looking at the list of Self-Paced Online Courses (the "Training Courses" link from the Help Center, direct link here) I'm not 100% sure if the Optimization Onramp covers Global Optimization Toolbox as well or if it focuses mainly or solely on Optimization Toolbox. But it may be of interest and/or use to you.