Search

By exploiting the computing power and local data of distributed clients, federated learning (FL) features ubiquitous properties such as reduction of communication overhead and preserving data privacy. In each communication round of FL, the clients update local models based on their own data and upload their local updates via wireless channels. However, latency caused by hundreds to thousands of communication rounds remains a bottleneck in FL. To minimize the training latency, this work provides a multi-armed bandit-based framework for online client scheduling (CS) in FL without knowing wireless channel state information and statistical characteristics of clients. Firstly, we propose a CS algorithm based on the upper confidence bound policy (CS-UCB) for ideal scenarios where local datasets of clients are independent and identically distributed (i.i.d.) and balanced. An upper bound of the expected performance regret of the proposed CS-UCB algorithm is provided, which indicates that the regret grows logarithmically over communication rounds. Then, to address non-ideal scenarios with non-i.i.d. and unbalanced properties of local datasets and varying availability of clients, we further propose a CS algorithm based on the UCB policy and virtual queue technique (CS-UCB-Q). An upper bound is also derived, which shows that the expected performance regret of the proposed CS-UCB-Q algorithm can have a sub-linear growth over communication rounds under certain conditions. Besides, the convergence performance of FL training is also analyzed. Finally, simulation results validate the efficiency of the proposed algorithms. [ABSTRACT FROM AUTHOR]

This paper considers a downlink nonorthogonal multiple access (NOMA) system where the source intends to transmit independent information to the users at target data rates under statistical channel state information. The outage balancing problem is studied with the issues of power allocation, decoding order selection, and user grouping being taken into account. Specifically, with regard to the max–min fairness criterion, we derive the optimal power allocation in closed form and prove the corresponding optimal decoding order for the elementary downlink NOMA system. By assigning a weighting factor for each user, the analytical results can be used to evaluate the outage performance of the downlink NOMA system under various fairness constraints. Furthermore, we investigate the case with user grouping, in which each user group can be treated as an elementary downlink NOMA system. The associated problems of intergroup power and resource allocation are solved. The implementation complexity issue of NOMA is also considered with focus on that caused by successive interference cancellation and user grouping. The complexity and performance tradeoff is analyzed by simulations, which provides fruitful insights for the practical application of NOMA. The simulation results substantiate our analysis and show considerable performance gain of NOMA when compared with orthogonal multiple access. [ABSTRACT FROM PUBLISHER]

Sensing-throughput tradeoff is involved in opportunistic spectrum access (OSA). At the physical layer it's expressed as the tradeoff between false alarm and misdetection, while at the MAC layer it's between maximizing the throughput of secondary users and reducing collisions with primary users. To balance both tradeoffs together drives the optimization of MAC frame structure for OSA. Since channel handoff results in a time overhead for MAC frame, it's first researched by modeling OSA dynamics in the paper. Three handoff cases and their probabilities are deduced, which lead to three application scenarios of MAC frame. Thus the throughput model of secondary network involving channel handoff is proposed. By balancing the misdetection and false alarm probabilities and maximizing the throughput of secondary network subject to the sensing quality and collision avoidance constraints, the optimal sensing time and frame duration are deduced as closed forms. The characteristics of the optimal MAC frame structure are researched as well. Theoretical and simulated results disclose the impacts of channel handoff and spectrum sensing on MAC frame structure and the achievable throughput of secondary network, which provide an insight for OSA design and improvement. [ABSTRACT FROM PUBLISHER]