gaplotpareto for 3 objective functions

is it possible to generate pareto front for 3 objective functions using GAPLOTPARETO in MATLAB? what about drawing first and third objective function pareto front instead of the 2 first objectives by using GAPLOTPARETO?

[Max]

I'm asking myself that same question. When looking at gaplotpareto, it's obvious, that the function is capable of producing 3D plots when treating an optimization problem of equivalent order, since the 'plot3' function and a variable called 'plot3d' appear inside the code.

However, the first comment line of the function says "GAPLOTPARETO Plots a Pareto front for first two objectives."

Maybe someone else can clarify this. Thanks!

Answer 1

[Alan Weiss]

Without guaranteeing anything, I think you can plot any two objectives using the syntax

 objectivesToPlot = [1,3]; % or whatever two objectives you want
 plotfn = @(options,state,flag)gaplotpareto(options,state,flag,objectivesToPlot);
 options = gaoptimset('PlotFcns',plotfn);

For three objective functions in a 3-D plot, try setting

 objectivesToPlot = [2 4 7]; % or whatever you want

Please let me know if it works for you. If so, maybe I should document it.

Alan Weiss

MATLAB mathematical toolbox documentation

[Max]

Thanks for your answer. Your suggestion does the trick for me. My actual problem was that I couldn't figure out how to set

 objectivesToPlot

correctly and hand it over to the plot function.

I think that adding an according section to the documentation of this function might be helpful to others as well.

EDIT: I've been using a version of

 gaplotpareto

that I modified to produce a plot that fits my needs. It works well for axes labels and so on but I can't seem to set

 hold on

for some reason. Does this have to do with the nature of the genetic algorithm repeatedly producing new populations (i.e. points to plot)? What I would like is all points from all generations being displayed so I get more of a pareto 'front' rather than a few scattered points.

[Alan Weiss]

I think that the code for gaplotpareto overwrites previous points on purpose. As the optimization progresses, the front may change, and you wouldn't want the old points anymore because they might not be on the front.

If you want more points on the front, use a larger population. Set the PopulationSize option to something larger (but be aware that this can slow the algorithm).

If you believe I have answered your questions helpfully, please accept my answer.

Alan Weiss

MATLAB mathematical toolbox documentation

[Max]

I figured out that increasing the population size would effectively lead to more of a 'front'. Thanks again.

[e_frog]

@Alan Weiss I had the EXACT same problem as Eshan and Max and your answer helped wonderfully! Thanks. Adding that code fragement to the documentation would be really nice!

I still have a followup question: How would I call @gaplotpareto after the optimization ended? I would like to plot the 3 subplots (obj1 over obj2, obj1 over obj3, obj2 over obj3) of my 3-dimensional problem (3 objectives).

[Alan Weiss]

I think that the topic Plot 3-D Pareto Front addresses this question.

Alan Weiss

MATLAB mathematical toolbox documentation

[Lucas Bruck]

In my case, the suggestion needed a minor fix:

options = gaoptimset(options,'PlotFcns',plotfn);