Approximate augmented Lagrangian functions and nonlinear semidefinite programs (Q2505379)

In this paper an approximate augmented Lagrangian approach is proposed for solving a nonlinear semidefinite program. The corresponding approximate Lagrangian function has the same order of smoothness as the functions in the original problem, which is a potential advantage in the design of optimization algorithms. After giving some basic properties of this approximate Lagrangian function such as monotonicity and convexity, the authors study necessary and sufficient conditions for approximate strong duality results. They also present conditions to obtain an approximate exact penalty representation in the framework of augmented Lagrangians. Concerning the convergence it is proven that under the Mangasarian Fromovitz condition, any limit point of a sequence of stationary points of approximate augmented Lagrangian problems is a KKT point of the original semidefinite program. Furthermore, under certain conditions, it is also shown that there exists a path of optimal solutions to approximate augmented Lagrangian problems converging to a solution of the semidefinite program.