1: \begin{abstract}

2: Asynchronous methods for solving systems of linear equations have

3: been researched since Chazan and Miranker's pioneering 1969 paper

4: on chaotic relaxation. The underlying idea of asynchronous methods

5: is to avoid processor idle time by allowing the processors to continue

6: to make progress even if not all progress made by other processors

7: has been communicated to them.

8: 

9: Historically, the applicability of asynchronous methods for solving linear equations 

10: was limited to certain restricted classes of matrices, such as diagonally dominant

11: matrices. Furthermore, analysis of these methods

12: focused on proving convergence in the limit. Comparison of the asynchronous

13: convergence rate with its synchronous counterpart and its scaling

14: with the number of processors were seldom studied, and are still not

15: well understood. 

16: 

17: In this paper, we propose a randomized shared-memory asynchronous method for general

18: symmetric positive definite matrices. We rigorously analyze the convergence

19: rate and prove that it is linear, and is close to that of the method's

20: synchronous counterpart if the processor count is not excessive relative

21: to the size and sparsity of the matrix. We also present an algorithm

22: for unsymmetric systems and overdetermined least-squares. Our work

23: presents a significant improvement in the applicability of asynchronous linear solvers

24: as well as in their convergence analysis, and suggests

25: randomization as a key paradigm to serve as a foundation for asynchronous

26: methods.

27: \end{abstract}

28: