1: \begin{abstract}

2: We introduce a reduced-communication distributed optimization scheme based on estimating

3: the solution to a proximal minimization problem. Our proposed setup involves a group of

4: agents coordinated by a central entity, altogether operating in a collaborative framework.

5: The agents solve proximal minimization problems that are hidden from the central

6: coordinator. The scheme enables the coordinator to construct a convex set within which the agents'

7: optimizers reside, and to iteratively refine the set every time that an agent is queried.

8: We analyze the quality of the constructed sets by showing their connections to the

9: $\epsilon$-subdifferential of a convex function and characterize their size. We prove convergence

10: results related to the solution of such distributed optimization problems and we devise a

11: communication criterion that embeds the proposed scheme in the Alternating Direction Method of

12: Multipliers (ADMM). The developed scheme demonstrates significant communication reduction when applied

13: to a microgrid setting.

14: \end{abstract}

15: