1: \begin{abstract}

2: Linearized alternating direction method of multipliers (ADMM) as an extension of

3: ADMM has been widely used to solve linearly constrained problems in signal processing,

4: machine leaning, communications, and many other fields.

5: Despite its broad applications in nonconvex optimization,

6: for a great number of nonconvex and nonsmooth objective functions,

7: its theoretical convergence guarantee is still an open problem.

8: In this paper, we propose a two-block linearized ADMM

9: and a multi-block parallel linearized ADMM

10: for problems with nonconvex and nonsmooth objectives.

11: Mathematically, we present that the algorithms can converge

12: for a broader class of objective functions under less strict assumptions

13: compared with previous works.

14: Furthermore, our proposed algorithm can update coupled variables in parallel

15: and work for less restrictive nonconvex problems,

16: where the traditional ADMM may have difficulties

17: in solving subproblems.

18: %More importantly, both algorithms can be applied to

19: %solve much more general problems,

20: %where the traditional ADMM, even with some theoretical convergence guarantee,

21: %may not be practical due to difficulties in solving subproblems.

22: 

23: \textbf{Keywords}:

24:  Linearized ADMM, nonconvex optimization, multi-block ADMM, parallel computation, proximal algorithm.

25: \end{abstract}

26: