1: \begin{abstract}

2: 

3: 

4: The approximation of tensors has important applications in various

5: disciplines, but it remains an extremely challenging task. It is

6: well known that tensors of higher order can fail to have best

7: low-rank approximations, but with an important  exception that best

8: rank-one approximations always exists. The most popular approach to

9: low-rank approximation is the alternating least squares (ALS)

10: method. The convergence of the alternating least squares algorithm

11: for the rank-one approximation problem is analysed in this paper. In

12: our analysis we are focusing on the global convergence and the rate

13: of convergence of the ALS algorithm. It is shown that the ALS method

14: can converge sublinearly, Q-linearly, and even Q-superlinearly. Our

15: theoretical results are illustrated on explicit examples.

16: 

17: \end{abstract}