On the rate of convergence of the conjugate gradient reset method with inaccurate linear minimizations

Recently Klessig and Polak have shown that a variation of the Polak-Ribière conjugate gradient reset algorithm converges
n
-step quadratically even if the one-dimensional search is not performed exactly. It is shown here that
n
-step quadratic convergence is also obtained for a modification of the Fletcher-Reeves conjugate gradient reset algorithm with inaccurate one-dimensional search.