A modified predictor-corrector method for linear programming (Q1318279)

The authors give a variant of the predictor-corrector method of \textit{M. Mizuno}, \textit{M. J. Todd} and \textit{S. Ye} [Technical Report No. 944, School of Operations Research and Industrial Engineering, Cornell University 1990] by modifying its corrector step so that the duality gap is as well reduced by a constant factor after this step. It is shown that the proposed method retains the \(O(\sqrt nL)\) iteration complexity and the local quadratic property.