What Is Optimization Toolbox?

fminuncY is a matrix that has the same number of rowsas there are dimensions in the problem. W =H*Y although H is not formed explicitly. fminuncuses Hinfo to compute the preconditioner. Theoptional parameters p1, p2, ... can be anyadditional parameters needed by hmfun. See“Avoiding Global Variables via Anonymous andNested Functions” on page 2-20 for informationon how to supply values for the parameters.Note 'Hessian' must be set to 'on' for Hinfoto be passed from fun to hmfun.HessPatternMaxPCGIterSee “Nonlinear Minimization with a Dense butStructured Hessian and Equality Constraints”on page 2-61 for an example.Sparsity pattern of the Hessian for finitedifferencing. If it is not convenient to computethesparseHessianmatrixH in fun, thelarge-scale method in fminunc can approximateH via sparse finite differences (of the gradient)provided the sparsity structure of H —i.e.,locations of the nonzeros—is supplied as thevalue for HessPattern. Intheworstcase,ifthestructure is unknown, you can set HessPatternto be a dense matrix and a full finite-differenceapproximation is computed at each iteration(this is the default). This can be very expensivefor large problems, so it is usually worth theeffort to determine the sparsity structure.Maximum number of PCG (preconditionedconjugate gradient) iterations (see “Algorithms”on page 8-88).8-84