Genetic algorithm' s options

Hi all,

I think to have a problem with the options for the ga solver. I set these options:

options = optimoptions('ga','MaxGenerations',1000,'PlotFcn',@gaplotbestf);

I believe that the solver ignores these options, do you know where the problem may be?

[Adam]

Depends what you do with those options and where your 'belief' stems from - i.e. is it a certainty? If not how do you know they aren't being used?

[Walter Roberson]

Are you passing those options to ga in the proper slot?

[Luca Riccardi]

This is how I pass them to the ga:

[X,FVAL,EXITFLAG,OUTPUT] = ga(@(variabiles) calibTM(variabiles,input_for_calib,Tpyrometers,ERR,Tpir_mod),8,[],[],[],[],lb,ub,[],[],options);

The slot should be correct.

[Walter Roberson]

That looks plausible.

What do you see that indicates to you that the options are being ignored?

[Luca Riccardi]

For example, the plot that should show the evolution of the fitness function value doesn't appear, and also the program is running since the day before yesterday and so for sure it created more than 1000 generations, which is the stopping criteria I selected.

[Star Strider]

Put a fprintf statement in ‘calibTM’ that prints the scalar output of the function either to your Command Window or to a file.

Also, posting or attaching ‘calibTM’ here could help.

[Luca Riccardi]

I already did it, the score of the fitness function keeps decreasing (I want to minimize it, so it is fine) but I want to stop the iterations some how.

[Star Strider]

I don’t know the default number of generations between the 'PlotFcn' calls, since that value is not documented and it isn’t possibly to query it. See if adding:

'PlotInterval',1

to your optimoptions argument list helps. That will plot after each generation.

[Luca Riccardi]

Nothing happens, but I guess that the problem is in the fitness function, since, trying with another one, it works correctly. Maybe it is too slow and so more than 1000 generations are required for converging.