How to use both pswplotbestf and pswplotranges for particleswarm?

I've noticed that having both 'pswplotranges' as 'OutputFcn' and 'pswplotbestf' as 'PlotFcn' in the particleswarm option enabled does not work. It appears that it's only possible to get either one of the plots from an optimization. Is there a way to get both? Thank you.

[Ashutosh]

The use of "pswplotranges” as the "OutputFcn” and "pswplotbestf” as the "PlotFcn" in the "particleswarm" options does not work is due to way the Particle Swarm Optimization algorithm handles the plotting.

The issue arises when using both options of "pswplotranges" as the "OutputFcn" and " pswplotbestf " as the "PlotFcn". These two functions are designed to generate different types of plots. "pswplotranges" generates a range plot that shows the range of particle positions at each iteration, while "pswplotbestf" generates a plot showing the best function value over iterations.

Since both functions are intended to generate plots, there is a conflict in how they are called and handled by the "particleSwarm" algorithm. The algorithm is designed to handle either an "OutputFcn" or a "PlotFcn", but not both simultaneously. As a result, when both functions are specified, it leads to errors or unexpected behavior.

Answer 1

[Alan Weiss]

The issue is that the pswplotranges output function was written incorrectly, and does not take into account that other plots might be happening at the same time. I will correct this deficiency in future documentation. Meanwhile, here is a corrected version along with a test script showing that the two plots can coexist. The four new lines in pswplotranges have % New comments at the ends.

Alan Weiss

MATLAB mathematical toolbox documentation

fun = @multirosenbrock;
nvar = 4; % A 4-D problem
lb = -10*ones(nvar,1); % Bounds to help the solver converge
ub = -lb;

options = optimoptions(@particleswarm,'OutputFcn',@pswplotranges,'PlotFcn',@pswplotbestf);
rng default % For reproducibility
[x,fval,eflag] = particleswarm(fun,nvar,lb,ub,options)

Optimization ended: relative change in the objective value 
over the last OPTIONS.MaxStallIterations iterations is less than OPTIONS.FunctionTolerance.

x = 1×4

0.9964 0.9930 0.9835 0.9681

fval = 3.4935e-04

eflag = 1

function stop = pswplotranges(optimValues,state)

stop = false; % This function does not stop the solver
switch state
    case 'init'
        fig = figure(); % New
        set(fig,'numbertitle','off','name','pranges') % New

        nplot = size(optimValues.swarm,2); % Number of dimensions
        for i = 1:nplot % Set up axes for plot
            subplot(nplot,1,i);
            tag = sprintf('psoplotrange_var_%g',i); % Set a tag for the subplot
            semilogy(optimValues.iteration,0,'-k','Tag',tag); % Log-scaled plot
            ylabel(num2str(i))
        end
        xlabel('Iteration','interp','none'); % Iteration number at the bottom
        subplot(nplot,1,1) % Title at the top
        title('Log range of particles by component')
        setappdata(gcf,'t0',tic); % Set up a timer to plot only when needed
    case 'iter'
        fig = findobj(0,'Type','figure','Name','pranges'); % New
        set(0,'CurrentFigure',fig); % New

        nplot = size(optimValues.swarm,2); % Number of dimensions
        for i = 1:nplot
            subplot(nplot,1,i);
            % Calculate the range of the particles at dimension i
            irange = max(optimValues.swarm(:,i)) - min(optimValues.swarm(:,i));
            tag = sprintf('psoplotrange_var_%g',i);
            plotHandle = findobj(get(gca,'Children'),'Tag',tag); % Get the subplot
            xdata = plotHandle.XData; % Get the X data from the plot
            newX = [xdata optimValues.iteration]; % Add the new iteration
            plotHandle.XData = newX; % Put the X data into the plot
            ydata = plotHandle.YData; % Get the Y data from the plot
            newY = [ydata irange]; % Add the new value
            plotHandle.YData = newY; % Put the Y data into the plot
        end
        if toc(getappdata(gcf,'t0')) > 1/30 % If 1/30 s has passed
          drawnow % Show the plot
          setappdata(gcf,'t0',tic); % Reset the timer
        end
    case 'done'
        % No cleanup necessary
end
end

function F = multirosenbrock(x)
% This function is a multidimensional generalization of Rosenbrock's
% function. It operates in a vectorized manner, assuming that x is a matrix
% whose rows are the individuals.

% Copyright 2014 by The MathWorks, Inc.

N = size(x,2); % assumes x is a row vector or 2-D matrix
if mod(N,2) % if N is odd
    error('Input rows must have an even number of elements')
end

odds  = 1:2:N-1;
evens = 2:2:N;
F = zeros(size(x));
F(:,odds)  = 1-x(:,odds);
F(:,evens) = 10*(x(:,evens)-x(:,odds).^2);
F = sum(F.^2,2);
end