We propose a novel Federated Edge Network Utility Maximization (FEdg-NUM) architecture for solving a large-scale distributed network utility maximization (NUM) problem. In FEdg-NUM, clients with private utilities communicate to a peer-to-peer network of edge servers. This represents a departure from the classical distributed NUM master-slave configuration and enables distributed computing harnessing local communications. Compared to a solution using cloud synchronization via Ring AllReduce, we prove that our federated edge computing model has shorter run-time in the presence of network congestion, thanks to its configuration and its ability to make progress in the presence of intermittent links. The paper studies its convergence and run-time performance both analytically and numerically, and illustrates several possible networking applications.

In federated learning, models are learned from users’ data that are held private in their edge devices, by aggregating them in the service provider’s “cloud” to obtain a global model. Such global model is of great commercial value in, e.g., improving the customers’ experience. In this paper we focus on two possible areas of improvement of the state of the art. First, we take the difference between user habits into account and propose a quadratic penalty-based formulation, for efficient learning of the global model that allows to personalize local models. Second, we address the latency issue associated with the heterogeneous training time on edge devices, by exploiting a hierarchical structure modeling communication not only between the cloud and edge devices, but also within the cloud. Specifically, we devise a tailored block coordinate descent-based computation scheme, accompanied with communication protocols for both the synchronous and asynchronous cloud settings. We characterize the theoretical convergence rate of the algorithm, and provide a variant that performs empirically better. We also prove that the asynchronous protocol, inspired by multi-agent consensus technique, has the potential for large gains in latency compared to a synchronous setting when the edge-device updates are intermittent. Finally, experimental results are provided that corroborate not only the theory, but also show that the system leads to faster convergence for personalized models on the edge devices, compared to the state of the art.

A futuristic infrastructure model is envisioned with distributed modules that can produce solar energy, have a storage system and provide services of lighting, electric-vehicle charging and communications. A stochastic model is formulated for the solar power production and overall consumption of power. Resource allocation in such a system translates to taking decisions in a foresighted manner while maximizing a social surplus with the goal of serving all power demands optimally. A stochastic dynamic programming approach is formulated for the same and a myopic policy is illustrated while discussing the trade-offs between serving competing demands.

We study uncoordinated random access over fading channels where each user independently decides whether to send a packet or not to a common receiver at any given time slot. Specifically, we develop an information theoretic formulation to characterize the overall system throughput. We consider two scenarios: classical slotted ALOHA, where no multiuser detection (MUD) capability is available and slotted ALOHA with MUD. In each case, in order to maximize the system throughput, we provide methods to obtain the optimal rates and channel activity probabilities using the user distances to the receiver (or, equivalently, their average signal to noise ratios) assuming a Rayleigh block fading channel. The results demonstrate that the newly proposed optimal rate selection solutions offer significant increase in the expected system throughputs compared to the “same rate to all users” approach commonly used in the literature. In addition to the overall throughput optimization, we also address the issue of fairness among users and propose approaches guaranteeing a minimum amount of individual throughput to each user, and design systems with limited individual outage probabilities for increased energy efficiency and reduced delay.

We describe a distributed framework for resource sharing problems that arise in communications, micro-economics, and various networking applications. In particular, we consider a hierarchical multi-layer decomposition for network utility maximization (ML-NUM), where functionalities are assigned to different layers. The proposed methodology creates solutions with central management and distributed computations to the resource allocation problems. In non-stationary environments, the technique aims to respond quickly to the dynamics of the network by decreasing delay by partially shifting the communication and computational burden to the network edges. Our main contribution is a detailed analysis under the assumption that the network changes are on the same time-scale as the convergence time of the algorithms used for local computations. Moreover, assuming strong concavity and smoothness of the users' objective functions, and under some stability conditions for each layer, we present convergence rates and optimality bounds for the ML-NUM framework. In addition, the main benefits of the proposed method are demonstrated with numerical examples.

In this paper, we consider a random access scheme over wireless fading channels based on slotted ALOHA where each user independently decides whether to send a packet or not to a common receiver at any given time slot. To characterize the system throughput, i.e., the expected sum- rate, an information theoretic formulation is developed. We consider two scenarios: classical slotted ALOHA where no multi-user detection (MUD) capability is available and slotted ALOHA with MUD. Our main contribution is that the optimal rates and channel activity probabilities can be characterized as a function of the user distances to the receiver to maximize the system throughput. In addition, we address the issue of fairness among the users and provide solutions, which guarantee a minimum amount of individual throughput.