How to define a custom plot function for an optimization

9 views (last 30 days)
Hi,
I want to define a custom plotfunction for my optimization. It should be a bit like the standard 'optimplotfval' plot function, but with a few extras. The information available is a bit unclear to me. It says that I have to define it just like the OutFnc, which I'm using too. If I'm using the same structure, I get an error message:
??? Error using ==> opt>objectiveconstraints/plotfun
Too many output arguments.
Error in ==> callAllOptimPlotFcns at 62
state(i) = feval(plotNames{i},x,optimvalues,'init',varargin{:});
Error in ==> nlconst>callOutputAndPlotFcns at 1032
stop = callAllOptimPlotFcns(plotfcns,xOrigShape,optimValues,state,varargin{:}) || stop;
Error in ==> nlconst at 559
[xOrigShape, optimValues, stop] = callOutputAndPlotFcns(outputfcn,plotfcns,caller,problemInfo, ...
Error in ==> fmincon at 724
[X,FVAL,LAMBDA,EXITFLAG,OUTPUT,GRAD,HESSIAN]=...
Error in ==> opt>objectiveconstraints at 260
[x,fval] = fmincon(problem);
Error in ==> opt at 50
[x,fval,aircraft] = objectiveconstraints(aircraft,problem);
I have defined my plot function in the same way as the OutFnc:
%Define function handles
objective=@(x)obj(aircraftstart,x);
constraint=@con;
outputfunction=@outfun;
plotfunction=@plotfun;
The function is nested inside the main function and looks like this:
function [] = plotfun(x,optimValues,state)
figure(20)
hold on
if optimValues.constrviolation<=1E-6
plot(optimValues.iteration,optimValues.fval,'g')
else
plot(optimValues.iteration,optimValues.fval,'r')
end
hold off
end
I don't have any output from the plot function, yet it still says I have too many output arguments. The options of fmincon are:
problem.options = optimset('Display','iter','MaxIter',1000,'OutputFcn',outputfunction,'FunValCheck','on','Algorithm','active-set','DiffMinChange',0.05,'DiffMaxChange',10.0,'PlotFcns',plotfunction,'TolX',1e-20);
Any thoughts?

Accepted Answer

Alan Weiss
Alan Weiss on 18 Aug 2011
Your line
function [] = plotfun(x,optimValues,state)
should be
function stop = plotfun(x,optimValues,state)
Inside your function you should have a line
stop = false;
By the way, you should not set TolX to anything less than a few eps, such as 4*eps. I would not set it to less than 1e-14 without a very good reason. You have such a large value of DiffMinChange that I think you should not set TolX at all, or perhaps set it to DiffMinChange/10.
  1 Comment
Jorrit
Jorrit on 19 Aug 2011
Thanks!
About the TolX, I have removed that now, doesn't seem like it's making a difference. I'm still struggling with the optimization setup and it was still there, because I just tried a lot of stuff.

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!