Genetic Algorithm (Plot Function)

Hi,

I set up an genetic algorithm for running a curve fitting process in order to identify the parameters (a,b,c) of a model equation. The model equation should later predict the experimental data depending on variables (x,y,z).

I used the sum of squares as my objective function:

sse = sum(power(ExperimentalData - ModelOutput,2));

Now I like to observe the actual function (ModelOutput) whiles the algorithm runs using the parameters (design variables) which the algorithm finds at each iteration.

Can someone help me to create such a Plot function which I could include in the GA options (gaoptimset)? I don`t really know which values are availible during the iterations and how I can access them. My initial idea was just to create a function file which includes my model function. This function file gets the parameters plus a range for the variables x,y,z (eg. x =[0:0.001:5]; y =[0:0.001:2]; z =[0:0.001:7])

and the experimental data as an input in order to visualize how the model prediction would look like compared to the experimental data.

How could a solution look like?

I´m looking forward to somebody helping me out with this problem.

Best regards Joe

[Joe]

Is there anyone who can help me?

Answer 1

[Cam Salzberger]

Hello Joe,

I understand that you are looking to plot the current output of the model as the genetic algorithm is running. I believe that you will find the 'PlotFcns' property, that can be set with "gaoptimset", to be the most useful. There are a variety of built-in plotting functions. Of them, I believe that the "Best individual" function (@gaplotbestindiv) is what you would like to see. You can also create a custom function, and pass its handle to that property. Custom functions should match the format shown here.

I hope this helps!

-Cam