MATLAB: Defining a stopping criterion for fmincon

fmincon

I am looking at an optimisation problem with the following objective function:

Here,

is some data and and

is a function which I want to find the values of X for which I is a minimum. Now I thought I would have to set the following:

options = optimset('TolFun',1e-3);

problem = createOptimProblem('fmincon','x0',X_old,'objective',fun,'lb',LB,'ub',UB,'options',options);

Is this the right thing to do?

Best Answer

Hi,

Yes, you can use optimset. You can directly use fmincon istead of createOptimProblem(but both work the same).

Hope this helps!

Related Solutions

MATLAB: Fmincon works but GA won’t work

In general, ga is not as reliable a solver as fmincon. So if you have smooth objective and constraints, you should use fmincon. See Table for Choosing a Solver.

Alan Weiss

MATLAB mathematical toolbox documentation

MATLAB: Stopping criterion in GA

alexaa - what you can do is compare, at each generation, the best found solution with the best one from the previous generation/iteration. If the absolute difference between the two is less than some tolerance, then they can be considered equal, and you just increment a counter. When that counter hits 100, exit the loop (I'm guessing that you will be using a for or while loop).

Try the following

 % initialize a local variable to keep track of the previous generation's best fitness
 prevGenBestFitness     = [];
 % initialize our counter
 numGensWithSameFitness = 0;
 % define a tolerance (note that this assumes real fitness scores, if
 % your fitness scores are integer, then this should be set to one or 
 % something similar)
 tol = 0.00001;
 % do the GA; this will be your for or while loop
 while true
     % do the GA stuff
     % save the best fitness from the current generation
     bestFitness = …;
     % compare this best with the previous best (we use ~empty to capture
     % the first iteration when there is nothing to compare)
     if ~isempty(prevGenBestFitness)
         % we can compare
         % NOTE I'm assuming that the best fitness from the current generation is
         % always as good or better than the fitness from the previous generation
         if abs(prevGenBestFitness-bestFitness)<tol
             % the two fitnesses are considered identical so increment the
             % counter
             numGensWithSameFitness = numGensWithSameFitness + 1;
             if numGensWithSameFitness==100
                 % 100 generations with same fitness, so exit
                 break;
             end
         else
             % the two fitnesses are not identical so reset the counter
             numGensWithSameFitness = 0;
         end
     end
     prevGenBestFitness = bestFitness;
 end

Note that the above does make certain assumptions about the fitness/score for your population and so may have to be adjusted to take into account a different scoring scheme (or data type).

Try the above and see what happens!

Best Answer

Related Solutions

MATLAB: Fmincon works but GA won’t work

MATLAB: Stopping criterion in GA

Related Question