1.5-\(Q\)-superlinear convergence of an exterior-point method for constrained optimization (Q2481366)

The authors provide an exterior-point method for constrained optimization problems with both inequality and equality constraints according to \[ \text{minimize }f(x), \] \[ \text{subject to }c_i(x)\geq 0,\;i= 1,\dots, p,\;g_j(x)= 0,\;j= 1,\dots,q, \] where \(f,c_i,g_j: \mathbb{R}^n\to \mathbb{R}\) are twice continuously differentiable functions. The method based on the nonlinear rescaling-augmented Lagrangian technique, which generalizes the modified barrier-augmented Langrangian method. Instead of finding the primal approximation and updating the Lagrange multipliers at each step, the exterior-point method performs one Newton step for solving the primal dual system. A special way to increase of the penalty-barrier parameters leads to a 1.5-\(Q\)-superlinear rate of convergence in a neighborhood of the solution assuming the standard second-order optimality condition.