Primal Dual Interior Point convergence.

177 Views Asked by At

I have developed a Primal Dual Interior point algorithm to solve linear inequality constrained Quadratic problems. But sometimes it cannot reduce the residual so as to satisfy the stop criteria. I mean the residual decreases and reaches some value (that I think depends on the scale of the matrices) and then it gets stuck at that value. How should I deal with this dilemma? I have studied an article about Predictor Corrector method, but it is noted in the article that sometimes this method does not converge. Is there any method that leads to definite convergence? And one more question,Is PDIPM even a good choice for solving quadratic problems? Is there any more advanced stop criteria other than primal and residual norm ,that varies with scale and size of variables ?