Sequential systems of linear equations method for general constrained optimization without strict complementarity (Q557781): Difference between revisions

The authors propose a new infeasible sequential systems of linear equations (SSLE) algorithm for solving an equality and inequality constrained optimization problem (P), based on an \(\mathit{l}_1 - \text\textit{l}_\infty\) exact penalty function and the extended active set identification technique. The initial point of the new algorithm may be any point in the \(\alpha\)-perturbation set of the feasible set. At each iteration, only two or three reduced linear equations with the same coefficients are solved to obtain the search direction. An automatic adjustment rule for the choice of penalty parameter is also incorporated in the new algorithm which ensures that the penalty parameter be updated only finitely many times. Under a linear independence condition, the algorithm globally converges to a KKT point of the problem (P). It is shown that the convergence rate of the new algorithm is one-step superlinear without strict complementarity and under a condition weaker than the strong second-order sufficiency condition. Many numerical experiments show the good properties of the new algorithm.