Abstract
This paper is concerned with improvement of the convergence rate of the gradient method. If the Hessian of a quadratic performance index has the maximum spectra M and the minimum spectram (M≥m≥0), the convergence rate of a gradient methods for this problem generally depends on the ratio M/m. In this paper, a variable transformation is introduced to make the ratio M/m smaller. Based on this transformation technique, the modified gradient method is developed. An example is given to show the effectiveness of the proposed method.