METHOD OF MULTIPLE STEEPEST DESCENT
多步最速下降法

In this paper we propose the method of multiple steepest descent. It is quad-ratic convergence. If the object function is f(x) =1/2 x~TAx - x~Tb,where A is an n×n symmetric positive definite matric with k different eigenvalues then afterk-time line or search the minimum point of f will be found. In this case we also note thatthis method is equivalent to the method of partial conjugate gradients.