Scheduling Policies for Federated Learning in Wireless Networks

By Howard H. Yang, Zuozhu Liu, Tony Q. S. Quek, H. Vincent Poor. 2019

590 citations recorded by the source corpus.

Motivated by the increasing computational capacity of wireless user equipments (UEs), e.g., smart phones, tablets, or vehicles, as well as the increasing concerns about sharing private data, a new machine learning model has emerged, namely federated learning (FL), that allows a decoupling of data acquisition and computation at the central unit. Unlike centralized learning taking place in a data center, FL usually operates in a wireless edge network where the communication medium is resource-constrained and unreliable. Due to limited bandwidth, only a portion of UEs can be scheduled for updates at each iteration. Due to the shared nature of the wireless medium, transmissions are subjected to interference and are not guaranteed. The performance of FL system in such a setting is not well understood. In this paper, an analytical model is developed to characterize the performance of FL in wireless networks. Particularly, tractable expressions are derived for the convergence rate of FL in a wireless setting, accounting for effects from both scheduling schemes and inter-cell interference. Using the developed analysis, the effectiveness of three different scheduling policies, i.e., random scheduling (RS), round robin (RR), and proportional fair (PF), are compared in terms of FL convergence rate. It is shown that running FL with PF outperforms RS and RR if the network is operating under a high signal-to-interference-plus-noise ratio (SINR) threshold, while RR is more preferable when the SINR threshold is low. Moreover, the FL convergence rate decreases rapidly as the SINR threshold increases, thus confirming the importance of compression and quantization of the update parameters. The analysis also reveals a trade-off between the number of scheduled UEs and subchannel bandwidth under a fixed amount of available spectrum.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering, Federated & Distributed Learning, ML Systems & Hardware

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

In Federated & Distributed Learning

In ML Systems & Hardware

Citations in the corpus

Age-Based Scheduling Policy for Federated Learning in Mobile Edge Networks (2019, background): “Nevertheless, because the wireless medium is resource-constrained, the AP can only select a subgroup of UEs for parameter updates in each communication round, and the choice of selection can largely affect the convergence rate of FL [19].”
On Safeguarding Privacy and Security in the Framework of Federated Learning (2019, background): “In a wireless setting, the communication links between the server and the clients are uncertain and imperfect and this effect needs to be carefully studied in the design of the FL system, especially in the large scale one [15].”
Communication-Efficient and Distributed Learning Over Wireless Networks: Principles and Applications (2020, background): “Especially, frequent and simultaneous computation and communication are impractical within the large-scale systems, in which resources management via client scheduling is essential for the identification, coordination, and diminution of the aforementioned stragglers [127]–[132].”
Client Selection and Bandwidth Allocation in Wireless Federated Learning Networks: A Long-Term Perspective (2020, background): “In [28], the convergence performance of FL under three basic scheduling policies, namely random, round-robin and proportional fair, is analyzed.”
A Joint Learning and Communications Framework for Federated Learning Over Wireless Networks (2019, methodology): “However, the work in [26] focused attention on three specific scheduling policies and, hence, did not address optimal user selection and resource allocation to optimize the FL performance.”
Federated Learning for Vehicular Internet of Things: Recent Advances and Open Issues (2020, background): “[60] analyze the effect of different scheduling policies on FL in wireless environments.”
A Tutorial on Ultrareliable and Low-Latency Communications in 6G: Integrating Domain Knowledge Into Deep Learning (2021, background): “Such a framework can avoid high communication over-heads and the security issue caused by collecting private training data from all the users [113].”
A Survey on Federated Learning: The Journey From Centralized to Distributed On-Site Learning and Beyond (2021, background): “[126] proposed a model for analyzing and characterizing the performance of FL.”
Federated-Learning-Enabled Intelligent Fog Radio Access Networks: Fundamental Theory, Key Techniques, and Future Trends (2020, background): “n etwork F unctIon V IrtuAlIzAtIon The concept of network function virtualization (NFV) is to virtualize the network functions decoupled from the underlying hardware .”
Convergence Time Optimization for Federated Learning Over Wireless Networks (2020, methodology): “The work in [13] derived an analytical model to characterize the training loss of FL in wireless networks and evaluated the effectiveness of different scheduling policies.”
Federated Learning With Quantized Global Model Updates (2020, background): “Active device selection schemes have been introduced to alleviate significant variability in FL systems, where a subset of devices share the resources and participate at each iteration of training [19–23].”
Joint Device Scheduling and Resource Allocation for Latency Constrained Wireless Federated Learning (2020, background): “These assumptions are widely used in the literature of convergence analysis for FL [19], [20], [25], [26], although the loss functions of some machine learning models (e.”
Edge Artificial Intelligence for 6G: Vision, Enabling Technologies, and Applications (2021, methodology): “…basically depends on the edge devices selection, packet errors in the uplink transmission, and model parameter partition, for which user scheduling [238], power control [88], batchsize selection [239], aggregation frequency control [240], and bandwidth allocation [241] were provided to improve the…”