Kybernetika 28 no. 4, 249-262, 1992

Computational experience with improved conjugate gradient methods for unconstrained minimization

Ladislav Lukšan

Abstract:

The paper contains a description of new restart procedures for the conjugate gradient methods and a numerical investigation of the influence of line search and scaling on their efficiency. Computational results obtained by means of 15 sufficiently difficult problems are given.

Classification:

49M37, 90C30, 65K05, 90-08