1: \begin{abstract}

2: %

3: We study a type of Riemannian gradient descent (RGD) algorithm,

4: designed through Riemannian preconditioning, for optimization on $\mrkk$---the set of $m\times n$ real matrices with a fixed rank $k$. 

5: % 

6: Our analysis is based on a quotient geometric view of $\mrkk$: 

7: % 

8: by identifying this set with the quotient manifold of a two-term product space $\prodspstar$ of matrices with full column rank via matrix factorization,

9: % 

10: we find an explicit form for the update rule of the RGD algorithm, which leads to a novel approach to analysing their convergence behavior in rank-constrained optimization. 

11: % 

12: We then deduce some interesting properties that reflect how RGD distinguishes from other matrix factorization algorithms such as those based on the Euclidean geometry. 

13: % 

14: In particular, we show that the RGD algorithm are not only faster than

15: Euclidean gradient descent but also do not rely on the balancing technique to

16: ensure its efficiency while the latter does.

17: % 

18: Starting from the novel results, we further show that this RGD algorithm

19: is guaranteed to solve matrix sensing and matrix completion problems with

20: linear convergence rate, under mild conditions related to the restricted positive definiteness property. % 

21: % 

22: Numerical experiments on matrix sensing and completion are provided to demonstrate these properties. 

23: \end{abstract}

24: