1: \begin{abstract}

2: PageRank is a well-known centrality measure for the web 

3: used in search engines, representing the importance of 

4: each web page. 

5: In this paper, we follow the line of recent research 

6: on the development

7: of distributed algorithms for computation of PageRank, 

8: where each page computes its own PageRank value by

9: interacting with pages connected over hyperlinks.

10: Our approach is novel in that it is based on a 

11: reinterpretation of PageRank, which leads us to a set 

12: of algorithms with exponential convergence rates. 

13: We first employ gossip-type randomization 

14: for the page selections in the update iterations.

15: Then, the algorithms are generalized 

16: to deterministic ones, allowing simultaneous updates

17: by multiple pages. Finally, based on these algorithms, 

18: we propose a clustering-based scheme, in which groups 

19: of pages make updates by locally interacting among themselves

20: many times to expedite the convergence. 

21: In comparison with other existing techniques, 

22: significant advantages can be exhibited in their convergence 

23: performance, as demonstrated via numerical examples using

24: real web data,

25: and also in the limited amount of communication

26: required among pages.

27: \end{abstract}

28: