1: \begin{abstract}                          Motivated by economic dispatch and linearly-constrained resource

2: 	allocation problems, this paper proposes a class of novel

3: 	$\distnewton$ algorithms that approximate the standard Newton

4: 	optimization method. We first develop the notion of an optimal edge

5: 	weighting for the communication graph over which agents implement

6: 	the second-order algorithm, and propose a convex approximation for

7: 	the nonconvex weight design problem.

8: 	% This weight design formulates to a nonconvex bilinear

9: 	% optimization, and we propose a convex approximation that

10: 	% is loosely based on completing the square to compute

11: 	% adequate solutions.

12: 	We next build on the optimal weight design to develop a

13: 	\distnewtondisc algorithm which converges linearly

14: 	to the optimal solution for economic dispatch problems with unknown

15: 	cost functions and relaxed local box constraints. For the full

16: 	box-constrained problem, we develop a \distnewtoncont algorithm

17: 	which is inspired by first-order saddle-point methods and rigorously

18: 	prove its convergence to the primal and dual optimizers. A main

19: 	property of each of these distributed algorithms is that they only

20: 	require agents to exchange constant-size communication messages,

21: 	which lends itself to scalable implementations. Simulations

22: 	demonstrate that the $\distnewton$ algorithms with our weight design

23: 	have superior convergence properties compared to existing weighting

24: 	strategies for first-order saddle-point and gradient descent

25: 	methods.

26: \end{abstract}