COMPARATIVE STUDY BETWEEN A NEW METHOD OF CONJUGATE ADDRESSES AND QUASI-NEWTON METHODS IN UNCONSTRAINED OPTIMIZATION

Abstract

This paper shows a comparative study between some Quasi-Newton methods and a new method of conjugate directions in unconstrained optimization, based on a modification of the condition of the secant. The new method of conjugate directions was studied by Guoyin Li et al, where gradient and target function information are used to define the conjugate condition. The Quasi-Newton methods used to perform the numerical comparison are BFGS and DFP. The numerical results show that in most tests, the proposed conjugate address method shows an improvement in execution time.