Is it possible to increase the performance by representing the values of the variables in each iteration of the loop?

I'm doing a genetic algorithm to solve the TSP problem and I'm trying to plot the result in each iteration, but I'm obtaining a bad performance. I would like to know how to improve the code to achieve better efficiency.

Is it possible to increase the performance by representing the values of the variables in each iteration of the loop?

  while iga<maxit
      
    %.......
    
    iga=iga+1; %      
    rte = pop(1,:); %Best Route
    
    subplot(1,2,1)
    plot(model.x(rte),model.y(rte),'r','Marker','o','MarkerEdgeColor','black', 'MarkerFaceColor','white','MarkerSize',12);
    title(sprintf('Total Distance = %1.2f, Iteration = %d',cost(1),iga)); 
    
    vxlabels = arrayfun(@(n) {sprintf('%d', n)}, (1:length(model.dt.Points))'); %Number of each point
    Hpl = text(model.x, model.y, vxlabels, 'FontWeight', 'bold', 'HorizontalAlignment',...
    'center', 'BackgroundColor', 'none','FontSize',10)
    
    subplot(1,2,2)
    plot((1:iga-1),minc(1:iga-1),(1:iga-1),meanc(1:iga-1));
    title('Convergence');
    
    drawnow limitrate;
    
    %.......
    
  end

[Walter Roberson]

Could you explain what you mean by "representing each iteration of the loop" ?

[Izan Segarra]

I'm plotting the value of minc and meanc (2 vectors) every loop iteraton and the performans is really bad.

[Walter Roberson]

If you are talking about graphics performance, then you could plot less often, or you could improve the code for doing the plotting.

But perhaps you are talking about the performance of your GA algorithm in the sense of it taking a long time per iteration, or in the sense of it taking a long time to make useful improvements in position?

[Izan Segarra]

I am talking about graphics performance.

Without representing the algorithm it takes 2 seconds, however with the representation it takes 110 seconds

I don't know how can I improve the code for doing the plot with better performance.

Answer 1

[Walter Roberson]

ax1 = subplot(1,2,1);
ax2 = subplot(1,2,2);
linehandle1 = line(ax1, nan, nan, 'r', 'Marker', 'o', 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'white', 'MarkerSize', 12);
Hpl = [];
title1 = title('Total Distance and iteration');
linehandle2a = line(ax2, nan, nan);
linehandle2b = line(ax2, nan, nan);
title(ax2, 'Convergence');

while iga<maxit
      
    %.......
    
    iga=iga+1; %      
    rte = pop(1,:); %Best Route
    
    set(linehandl1, 'XData', model.x(rte), 'YData', model.y(rte));
    set(title1, 'string', sprintf('Total Distance = %1.2f, Iteration = %d',cost(1),iga)); 
    
    vxlabels = arrayfun(@(n) {sprintf('%d', n)}, (1:length(model.dt.Points))'); %Number of each point
    if ~isempty(Hpl); delete(Hpl); end
    Hpl = text(model.x, model.y, vxlabels, 'FontWeight', 'bold', 'HorizontalAlignment',...
    'center', 'BackgroundColor', 'none','FontSize',10)
    
    set(linehandle2a, 'XData', 1:iga-1, 'YData', minc(1:iga-1));
    set(linehandle2b, 'XData', 1:iga-1, 'YData', meanc(1:iga-1));
    
    drawnow limitrate;
    
    %.......
    
end

Although it might in theory be more efficient to update all of the text objects rather than creating new ones, there is one text object per label and so changing them all becomes a bit messier. It is possible even without loop, but it involves some advanced features.

[Izan Segarra]

Thanks for your answer!

I would like to learn how to update text objects instead of creating new ones to improve efficiency. Where could I learn more about this topic?

[Walter Roberson]

Is it the case that model.x and model.y do not change, and that rte is a vector that is consistent length -- that rte is just giving the order to use the values in? This is important for the question of efficiently updating the labels.

[Izan Segarra]

Exactly,the labels are always the same. Fuerthermore model.x and model.y never change and the vector rte always has the same length.

[Walter Roberson]

ax1 = subplot(1,2,1);
ax2 = subplot(1,2,2);
linehandle1 = line(ax1, nan, nan, 'r', 'Marker', 'o', 'MarkerEdgeColor', 'black', 'MarkerFaceColor', 'white', 'MarkerSize', 12);
Hpl = text(model.x, model.y, {}, 'FontWeight', 'bold', 'HorizontalAlignment',...
    'center', 'BackgroundColor', 'none','FontSize',10);
title1 = title('Total Distance and iteration');
linehandle2a = line(ax2, nan, nan);
linehandle2b = line(ax2, nan, nan);
title(ax2, 'Convergence');

while iga<maxit
      
    %.......
    
    iga=iga+1; %      
    rte = pop(1,:); %Best Route
    
    set(linehandl1, 'XData', model.x(rte), 'YData', model.y(rte));
    set(title1, 'string', sprintf('Total Distance = %1.2f, Iteration = %d',cost(1),iga)); 
    
    vxlabels = num2str(rte(:));
    set(Hpl, {'String'}, vxlabels);
    
    set(linehandle2a, 'XData', 1:iga-1, 'YData', minc(1:iga-1));
    set(linehandle2b, 'XData', 1:iga-1, 'YData', meanc(1:iga-1));
    
    drawnow limitrate;
    
    %.......
    
end

[Izan Segarra]

Thank you very much for your answer!