1: \begin{abstract}

2: 

3: Eigenvectors of data matrices play an important role in many computational

4: problems, ranging from signal processing to machine learning and control.

5: For instance, algorithms that compute positions of the nodes of a

6: wireless network on the basis of pairwise distance measurements require 

7: a few leading eigenvectors of the distances matrix.

8: While eigenvector calculation is a standard topic in numerical linear

9: algebra, it becomes challenging under severe communication or 

10: computation constraints, or in absence of central scheduling.

11: In this paper we investigate the possibility of computing the leading

12: eigenvectors of a large data matrix through gossip algorithms.

13: 

14: The proposed algorithm amounts to iteratively multiplying 

15: a vector by independent random

16: sparsification of the original matrix and averaging the resulting

17: normalized vectors. This can be viewed as a generalization of

18: gossip algorithms for consensus, but the resulting dynamics is significantly 

19: more intricate. Our analysis is based on controlling the

20: convergence to stationarity of the associated Kesten-Furstenberg Markov chain.

21: 

22: \end{abstract}

23: