1: \begin{abstract}

2:   The so-called Burer-Monteiro method is a well-studied technique for solving

3:   large-scale semidefinite programs (SDPs) via low-rank factorization. The main idea is to solve rank-restricted, albeit \emph{non}-convex, surrogates instead of the original

4:   SDP. Recent works have shown that, in an important class of SDPs with an

5:   elegant geometric structure, one can find globally optimal solutions to the

6:   SDP by finding rank-deficient second-order critical points of an unconstrained

7:   Riemannian optimization problem. Hence, in such problems, the Burer-Monteiro approach can

8:   provide a scalable and reliable alternative to interior-point methods that

9:   scale poorly.

10:   Among various Riemannian optimization methods proposed, block-coordinate

11:   minimization (BCM) is of particular interest due to its simplicity.

12:   \citet{ErdogduBCM2018} in their recent work proposed BCM for problems over the

13:   Cartesian product of unit spheres and 

14:   provided global convergence rate estimates for the algorithm.

15:   This report extends the BCM algorithm and the global convergence

16:   rate analysis of

17:   \citet{ErdogduBCM2018} from problems over the Cartesian product of unit spheres

18:   to the Cartesian product of Stiefel manifolds. The latter more general setting

19:   has important applications such as synchronization over the special

20:   orthogonal (SO) and special Euclidean (SE) groups.

21: \end{abstract}