1: \begin{abstract}

2: A distributed algorithm for least mean square (LMS)

3: can be used in distributed signal estimation and 

4: in distributed training for multivariate regression models.

5: The convergence speed of an algorithm is a critical factor because 

6: a faster algorithm requires less communications overhead and

7: it results in a narrower network bandwidth.

8: The goal of this paper is to present that use of Chebyshev periodical 

9: successive over-relaxation (PSOR) can accelerate distributed LMS algorithms 

10: in a naturally manner. 

11: The basic idea of Chbyshev PSOR 

12: is to introduce 

13: index-dependent PSOR factors that control the spectral radius of a matrix governing the convergence 

14: behavior of the modified fixed-point iteration. Accelerations of convergence speed are 

15: empirically confirmed in a wide range of networks, such as 

16: known small graphs (e.g., Karate graph),  and random graphs, such as

17: Erd\"os-R\'enyi (ER) random graphs and  Barab\'asi-Albert random graphs.

18: \end{abstract}

19: