how to manage fzero output

This is a beginner's question.

I am using particle swarm optimization (PSO). The fzero function analyzes the values assigned by PSO in the objective function. Sometimes, however, the values assigned by PSO violate the input limits of fzero, which interrupts the process.

Is there a way to manage the output of fzero through optimoptions or optimset?

Answer 1

[John D'Errico]

I'd argue you are thinking about this the wrong way.You see that fzero has a problem, so you think you need to do something to fzero. But that is wrong, the wrong approach.

You CAN constrain the parameters in the PSO however! Nothing stops you from doing that. Why do you need to use optimoptions? (Which would be of absolutely no use anyway.) Note that particleswarm does allow bound constraints on the parameters.

help particleswarm

 particleswarm - Particle swarm optimization
    This MATLAB function attempts to find a vector x that achieves a local
    minimum of fun.

    Syntax
      x = particleswarm(fun,nvars)
      x = particleswarm(fun,nvars,lb,ub)
      x = particleswarm(fun,nvars,lb,ub,options)
      x = particleswarm(problem)
      [x,fval,exitflag,output,points] = particleswarm(___)

    Input Arguments
      fun - Objective function
        function handle | function name
      nvars - Number of variables
        positive integer
      lb - Lower bounds
        [] (default) | real vector or array
      ub - Upper bounds
        [] (default) | real vector or array
      options - Options for particleswarm
        options created using optimoptions
      problem - Optimization problem
        structure

    Output Arguments
      x - Solution
        real vector
      fval - Objective value
        real scalar
      exitflag - Algorithm stopping condition
        integer
      output - Solution process summary
        structure
      points - Final swarm positions and objective function values
        structure

    Examples
      openExample('globaloptim/MinimizeaSimpleFunctionExample')
      openExample('globaloptim/MinimizeaSimpleFunctionwithBoundsExample')
      openExample('globaloptim/MinimizeUsingNondefaultPSwarmOptionsExample')
      openExample('globaloptim/ExaminetheSolutionProcessExample')

    See also ga, patternsearch, Optimize

    Introduced in Global Optimization Toolbox in R2014b
    Documentation for particleswarm
       doc particleswarm

If you are using some other tool, and you never told us, then you are on your own, since we cannot know what the capabilities of some unknown tool would be.

Finally, you have code that calls fzero. Surely you can pretest the parameters just before you call fzero. If the parameter would cause a problem, then do not call fzero, and just return some large value for the objective for that set of parameters. Again, there is absolutely no problem. All it takes is an if statement to test the value, BEFORE fzero is ever called.

And even if all of the above ideas don't seem to work for you (which they absolutely must, but assuming they don't) then you could call fzero using a try/catch clause to catch the times when fzero would fail.

Do you see that instead of looking at fzero, you need to think about how to deal with the problem before fzero ever sees it as an issue.

I've outlined 4 different solutions you can use. One of them will work for you. Actually, ALL of them should work.