Rate of Convergence of an Improved Reduced Gradient Method

An improved reduced gradient method was proposed in [4] to solve the nonlinear programming (p) with linear constraints: (p) f(x) R={x∈E~n|AX=b,x≥0} b∈E~m. In this paper we introduce parameters ρ_k which is the skill used in [5] to the algorithm of [4] to obtain a reduced gradient method which is linearly convergent under the conditions of R being non-degenerate, f being second-order continuously differentiable and strong convex.