1: \begin{abstract}

2: In this paper, we discuss the convergence and acceleration of

3: the regularized alternating least square (RALS) algorithm

4: for tensor approximation.

5: We analyze the global convergence based on

6: the Kurdyka-{\L}ojasiewicz inequality.

7: Moreover, we show that the RALS algorithm has a linear local convergence rate.

8: Furthermore, we propose an accelerated version for this regularized algorithm.

9: The numerical experiments verify the linear local convergence rate of the RALS algorithm and demonstrate a faster convergence rate for the accelerated version  in comparison to both the standard and regularized alternating least squares algorithms.

10: \end{abstract}

11: