1: \begin{abstract}

2: Most existing work uses dual decomposition and first-order methods

3: to solve Network Utility Maximization (NUM) problems in a

4: distributed manner, which suffer from slow rate of convergence

5: properties. This paper develops an alternative distributed

6: Newton-type fast converging algorithm for solving NUM problems with

7: self-concordant utility functions. By using novel matrix splitting

8: techniques, both primal and dual updates for the Newton step can be

9: computed using iterative schemes in a decentralized manner. We propose a stepsize rule and provide a distributed procedure to compute it in finitely many iterations. The key feature of our direction and stepsize computation schemes is that both are implemented using the same distributed information exchange mechanism employed by first order methods. We show that even when the

10: Newton direction and the stepsize in our method are computed within

11: some error (due to finite truncation of the iterative schemes), the

12: resulting objective function value still converges superlinearly in terms of primal iterations to

13: an explicitly characterized error neighborhood. Simulation results

14: demonstrate significant convergence rate improvement of our

15: algorithm relative to the existing first-order methods based on dual

16: decomposition.

17: \end{abstract}

18: