1: \begin{abstract}

2: In the Euclidean setting the proximal  gradient method and its accelerated variants are a class of efficient algorithms for  optimization problems with decomposable objective.

3: \whfirrev{}{In this paper, we develop a Riemannian proximal gradient method (RPG) and its accelerated variant (ARPG)  for similar problems but constrained on a manifold. The global convergence of RPG \whsecrev{}{is} established under mild assumptions, and the  $O(1/k)$ is also derived for RPG based on the notion of retraction convexity. If  assuming the objective function obeys the Rimannian Kurdyka-\L ojasiewicz (KL) property, it is further shown that the sequence generated by RPG converges to a single stationary point.

4:  As in the Euclidean setting, local convergence rate  can be established if the objective function satisfies the Riemannian KL property with an exponent. 

5:  Moreover, we \whsecrev{}{show} that the restriction of a semialgebraic function onto  the Stiefel manifold  satisfies the Riemannian KL property, which covers for example the well-known sparse PCA problem. }

6: %Empirical performance comparisons show that the proposed RPG and ARPG are competitive with existing ones.

7: \whsecrev{}{Numerical experiments on  random and synthetic data are conducted to test  the performance of the  proposed RPG and ARPG.}

8: \end{abstract}

9: