Efficient Algorithm for Unconstrained Optimization Problems ——A Positive Definite Update from Negative Broyden Family

Abstract A computational implementation of Broyden family of variable metric methods was presented, in which the parameter θ had distinct valurs at various iterations and θ was made less than zero as far as possible Numerical resultsof solving some test problems by the algorithms with a few schemes for selecting θand the BFGS method were given for comparison to demonstrate that the algorithmis efficient for unconstrained optimization problems