How to plot an optimization in logarithmic scale?

The following code is expected to plot the output of Rosenbrock's function against number of iterations ( for the sake of the problem, don't be concerned with the quality of the source code).

 function value = banana(x0) 
    a = x0(1); 
    b = x0(2); 
    x = x0(3); 
    y = x0(4); 
    value = (1 - x + a)^2 + 100 * (y - b * (x - a)^2)^2; 
 end  

a = int8(4 * rand()) / 2; 
b = int8(4 * rand()) / 2; 

[x1, y1] = random(a, b); 
[x2, y2] = random(a, b);
[x3, y3] = random(a, b); 
[x4, y4] = random(a, b); 

x = [x1; x2; x3; x4]; 
y = [y1; y2; y3; y4]; 

save values.mat; 

for i = 1 
    x0 = [a, b, x(i), y(i)]; 

      options = optimset('PlotFcns',  { @optimplotfval });     
      [solution_point, fval,exitflag,output] = fminsearch(@banana,double(x0),options); 

      old_x=[x(i),y(i)]; 
      new_x_1 = linspace(-2,2,51); 
      new_x_2 = linspace(-2,2,51); 

      for j=1:51 
          if(new_x_1(j)>= old_x(1)) 
              new_x_1(j)=old_x(1); 
              break; 
          end 
      end

      for j=1:51
          if(new_x_2(j)>= old_x(2)) 
              new_x_2(j)=old_x(2); 
              break; 
          end 
      end
  end

My questions are,

(1) Is this plot really showing the output of Rosenbrock's function against number of iterations?

(2) The plot shows a sudden slump in the output of the function. I need to show this plot in logarithmic form, so that a smooth curve is plotted. How can I do that?

Answer 1

[Walter Roberson]

For the log plot:

edit optimplotfval

now Save As into your own directory as optimplotlogfval.m . Close the original optimplotfval.m (so you do not accidentally modify it.) Then in the optimplotlogfval.m version, alter the "function" statement to use optimplotlogfval instead of optimplotfval as the name. Then go into the plotscalar function there and for the iteration == 0 case, after the ylabel() call, add

set(gca, 'YScale', 'log');

Make the same change for iteration == 0 in the plotvector function.

Saving all of that, go back to your code and change

        options = optimset('PlotFcns',  { @optimplotfval });

to

        options = optimset('PlotFcns',  { @optimplotlogfval });

[Obinna]

Would you still recommend to do this now? plotscalar seems to have changed from that to "thePlot", so I don't know if these intructions work for 2025 MATLAB. Would it be possible to give another instructional on how to do it currently? Thank you