1: \begin{abstract}

2: 	We consider solving high-order semidefinite programming (SDP) relaxations of nonconvex polynomial optimization problems (POPs) that often admit degenerate rank-one optimal solutions. 

3: 	Instead of solving the SDP alone, we propose a new algorithmic framework that blends \emph{local search} using the nonconvex POP into \emph{global descent} using the convex SDP. In particular, we first design a globally convergent \emph{inexact} projected gradient method (\ipgm) for solving the SDP that serves as the backbone of our framework. We then accelerate \ipgm\xspace by taking long, but \emph{safeguarded}, rank-one steps generated by fast nonlinear programming algorithms. We prove that the new framework is still globally convergent for solving the SDP. To solve the  \ipgm\xspace subproblem of projecting a given point onto the feasible set of the SDP, we design a two-phase algorithm with phase one using a symmetric Gauss-Seidel based accelerated proximal gradient method (sGS-APG) to generate a good initial point, and phase two using a modified limited-memory BFGS (\lbfgs) method to obtain an accurate solution. We analyze the convergence for both phases and establish a novel global convergence result for the modified \lbfgs\xspace that does not require the objective function to be twice continuously differentiable. We conduct numerical experiments for solving second-order SDP relaxations arising from a diverse set of POPs. Our framework demonstrates state-of-the-art efficiency, scalability, and robustness in solving degenerate rank-one SDPs to high accuracy, even in the presence of millions of equality constraints.

4: \end{abstract}

5: