A Alternating Direction Method For Quadratic Semidefinite Programming

The quadratic semidefinite programming(abbreviated as QSDP) is a branch of semidefinite programming(abbreviated as SDP). In this thesis, we mainly consider one kind of QSDP problem. We will meet a difficulty if using traditionary augmented Lagrangian method to solve this kind of QSDP problem:minimizing augmented La-grangian function is hard to be computed. Inspired by [11] and [26], we introduce an alternating direction augmented Lagrangian method to solve this kind of QSDP prob-lem. In this thesis, we prove the convergence of this alternating direction augmented Lagrangian method and by presenting some numerical experiments we show its conver-gence.