THE SEQUENTIAL UNCONSTRAINED MINIMIZATION TECHNIQUE FOR NONLINEAR PROGRAMMING. ALGORITHM II. OPTIMUM GRADIENTS BY FIBONACCI SEARCH

Abstract : The algorithm has been revised to incorporate a more efficient technique for computing the minimum of a function of a specified vector, a computation required in each iteration of the optimumgradient method. The new technique is an adaptation of a Fibonacci-gradient method. The new technique is an adaptation of a Fibonacci previously used and results in a recuction in total problem solution time of almost one half. A new normalized final-convergence criterion that does not depend on the magnitude of the optimum solution value is given. The detailed computer solution of a change-constrained linear programming problem illustrates the typical convergence characteristics of the method. The remainder of the paper is a concise and simplified review of all the method's important computational aspects. (Author)