Asymptotic convergence in a generalized predictor–corrector method

The asymptotic convergence properties of a generalized predictor–corrector method are analyzed. This method is based on making a sequence of corrections to the primal–dual affine scaling (predictor) direction. It is shown that a method making r corrections to a predictor direction has the Q-order convergence of order r + 2. It is also shown that asymptotically the problem can be solved by only computing corrections to the predictor direction.