Comment by ProfGrimmer

My paper gives guarantees for any smooth convex minimization problem. The semidefinite programming is needed to solve a meta-optimization problem in the analysis of these generic problems. It's a tool rather than the target.