A new conic method for unconstrained minimization

AbstractA new algorithm for unconstrained minimization is presented which is based on a conic model. The algorithm converges in (n + 1) iterations on conic functions of n variables and it does not require evaluation or estimation of the matrix of second partial derivatives. Numerical results on many standard general test functions indicate that the algorithm is very robust and superior in function evaluations and number of iterations to the corresponding quadratic method