A modified conjugate gradient method using multi-step in unconstrained optimization

Abstract EN

In this paper we proposed a new HS type conjugate gradient method by using two kind of modification, first derived a new non-quadratic model second using structure of the memoryless BFGS quasi-Newton method.

The new proposed method always generates a descent condition.

We give a sufficient condition for the global converges of the proposed general method.

Finally, some numerical results are also reported.