No abstract

Federated learning is a recent approach for distributed model training without sharing the raw data of clients. It allows model training using the large amount of user data collected by edge and mobile devices, while preserving data privacy. A challenge in federated learning is that the devices usually have much lower computational power and communication bandwidth than machines in data centers. Training large-sized deep neural networks in such a federated setting can consume a large amount of time and resources. To overcome this challenge, we propose a method that integrates model pruning with federated learning in this paper, which includes initial model pruning at the server, further model pruning as part of the federated learning process, followed by the regular federated learning procedure. Our proposed approach can save the computation, communication, and storage costs compared to standard federated learning approaches. Extensive experiments on real edge devices validate the benefit of our proposed method.

No abstract

According to the ongoing IEEE 802.11ac amendment, the wireless network is about to embrace the gigabit-per-second raw data rate. Compared with previous IEEE standards, this significant performance improvement can be attributed to the novel physical and medium access control (MAC) features, such as multi-user multiple-input multiple-output transmissions, the frame aggregation, and the channel bonding. In this paper, we first briefly survey the main features of IEEE 802.11ac, and then, we evaluate these new features in a fully connected wireless mesh network using an analytic model and simulations. More specifically, the performance of the MAC scheme defined by IEEE 802.11ac, which employs the explicit compressed feedback (ECFB) mechanism for the channel sounding, is evaluated. In addition, we propose an extended request-to-send/clear-to-send scheme that integrates the ECFB operation to compare with the IEEE 802.11ac-defined one in saturated conditions. The comparison of the two MAC schemes is conducted through three spatial stream allocation algorithms. A simple but accurate analytical model is derived for the two MAC schemes, the results of which are validated with simulations. The observations of the results not only reveal the importance of spatial stream allocations but also provide insight into how the newly introduced features could affect the performance of IEEE 802.11ac-based wireless mesh networks.

Future mobile networks will exploit unlicensed spectrum to boost capacity and meet growing user demands costeffectively. The 3GPP has recently defined a Licensed-Assisted Access (LAA) scheme to enable global Unlicensed LTE (U-LTE) deployment, aiming at (i) ensuring fair coexistence with incumbent WiFi networks, i.e., impacting on their performance no more than another WiFi device, and (ii) achieving superior airtime efficiency as compared to WiFi. In this paper we show the standardized LAA fails to simultaneously fulfill these objectives, and design an alternative orthogonal (collision-free) listen-beforetalk coexistence paradigm that provides a substantial improvement in performance, yet imposes no penalty on existing WiFi networks. We derive two LAA optimal transmission policies, ORLA and OLAA, that maximize LAA throughput in both asynchronous and synchronous (i.e., with alignment to licensed anchor frame boundaries) modes of operation, respectively. We present a comprehensive performance evaluation through which we demonstrate that, when aggregating packets, IEEE 802.11ac WiFi can be more efficient than 3GPP LAA, whereas our proposals can attain 100% higher throughput, without harming WiFi. We further show that long U-LTE frames incur up to 92% throughput losses on WiFi when using 3GPP LAA, whilst ORLA/OLAA sustain >200% gains at no cost, even in the presence of non-saturated WiFi and/or in multi-rate scenarios.