1: \begin{abstract}

2:   In this paper we consider a recently developed distributed optimization

3:   algorithm based on gradient tracking. We propose a system theory framework

4:   to analyze its structural properties on a preliminary, quadratic optimization

5:   set-up.  Specifically, we focus on a scenario in which agents in a static

6:   network want to cooperatively minimize the sum of quadratic cost functions.

7:   We show that the gradient tracking distributed algorithm for the investigated

8:   program can be viewed as a sparse closed-loop linear system in which

9:   the dynamic state-feedback controller includes consensus matrices and optimization

10:   (stepsize) parameters. The closed-loop system turns out to be not completely

11:   reachable and asymptotic stability can be shown restricted to a proper

12:   invariant set. 

13:   %

14: %

15:   Convergence to the global minimum, in turn, can be obtained only by 

16:   means of a proper initialization. The proposed system interpretation 

17:   of the distributed algorithm provides also additional insights on other 

18:   structural properties and possible design choices that are discussed in 

19:   the last part of the paper as a starting point for future developments. 

20:   %

21:   %

22:   %

23:   %

24:   %

25:   %

26:   %

27:   %

28:   %

29:   %

30: \end{abstract}

31: