1: \begin{abstract}

2:   In this paper, we study \emph{distributed big-data nonconvex} optimization in

3:   multi-agent networks.  We consider the (constrained) minimization of the sum

4:   of a smooth (possibly) nonconvex function, i.e., the agents' sum-utility, plus

5:   a convex (possibly) nonsmooth regularizer. Our interest is in big-data

6:   problems wherein there is a large number of variables to optimize. If treated

7:   by means of standard distributed optimization algorithms, these large-scale

8:   problems may be intractable, due to the prohibitive local computation and

9:   communication burden at each node.  We propose a novel distributed solution

10:   method whereby at each iteration agents optimize and then communicate  (in an uncoordinated fashion)

11:   only a subset of their decision variables. To deal with   non-convexity

12:   of the cost function, the novel scheme hinges on Successive Convex

13:   Approximation (SCA) techniques coupled with i) a tracking mechanism

14:   instrumental to locally estimate gradient averages; and ii) a novel

15:   \emph{block-wise} consensus-based protocol to perform local block-averaging

16:   operations and gradient tacking. Asymptotic convergence to stationary

17:   solutions of the nonconvex problem is established. Finally, numerical results

18:   show the effectiveness of the proposed algorithm and highlight how the block

19:   dimension impacts on the communication overhead and practical convergence

20:   speed. 

21: \end{abstract}

22: