What Is Optimization Toolbox?

Typical Problems and How to Deal with ThemTroubleshooting (Continued)Problemfminunc produceswarning messages andseems to exhibit slowconvergence near thesolution.Sometimes anoptimization problemhas values of x forwhich it is impossible toevaluate the objectivefunction fun or thenonlinear constraintsfunction nonlcon.The function that isbeing minimized hasdiscontinuities.RecommendationIf you are not supplying analytically determined gradients and thetermination criteria are stringent, fminunc often exhibits slowconvergence near the solution due to truncation error in the gradientcalculation. Relaxing the termination criteria produces faster,although less accurate, solutions. For the medium-scale algorithm,another option is adjusting the finite-difference perturbation levels,DiffMinChange and DiffMaxChange, which might increase theaccuracy of gradient calculations.Place bounds on the independent variables or make a penaltyfunction to give a large positive value to f and g when infeasibility isencountered. For gradient calculation, the penalty function should besmooth and continuous.The derivation of the underlying method is based upon functionswith continuous first and second derivatives. Some success mightbe achieved for some classes of discontinuities when they do notoccur near solution points. One option is to smooth the function.For example, the objective function might include a call to aninterpolation function to do the smoothing.Or, for the medium-scale algorithms, you can adjust thefinite-difference parameters in order to jump over smalldiscontinuities. The variables DiffMinChange and DiffMaxChangecontrol the perturbation levels for x used in the calculation offinite-difference gradients. The perturbation, ,isalwaysintherange DiffMinChange < Dx < DiffMaxChange.2-99