1: \begin{abstract}

2: In this note we take a new look at the local convergence of alternating optimization methods for low-rank matrices and tensors. Our abstract interpretation as sequential optimization on moving subspaces yields insightful reformulations of some known convergence conditions that focus on the interplay between the contractivity of classical multiplicative Schwarz methods with overlapping subspaces and the curvature of low-rank matrix and tensor manifolds. While the verification of the abstract conditions in concrete scenarios remains open in most cases, we are able to provide an alternative and conceptually simple derivation of the \changed{asymptotic convergence rate} of the two-sided block power method of numerical algebra for computing the dominant singular subspaces of a rectangular matrix. This method is equivalent to an alternating least squares method applied to a distance function. %{\color{red} Some implications regarding ALS for low-rank tensor approximation in tensor train format are also given.} 

3: The theoretical results are illustrated and validated by numerical experiments.

4: 

5: \medskip

6: 

7: \noindent

8: \textit{Keywords:} ALS, nonlinear Gauss--Seidel method, low-rank approximation, local convergence 

9: \end{abstract}

10: