2019 · 17 citations · 18 references
ComputingComputer ScienceComputer SystemsDistributed SystemsMachine LearningWireless CommunicationsDistributed SensingFederated LearningInterference ManagementSensor NetworksSpectrum ManagementWireless Sensor NetworksAccuracyBandwidthBandwidth OverheadCitizens Broadband Radio ServiceEnvironmental SensorsFederated Spectrum AccessSpatial DiversitySpectrum Sensing
The next-generation spectrum access system (SAS) for the Citizens Broadband Radio Service band is equipped with environmental sensors (ESCs) to detect the presence of non-informed incumbent users, which allows the SAS to dynamically reassign spectrum resource for low privilege users to avoid interference. However, the performance of existing single-node detection model is limited by the sensor’s geo-locations; whereas a naive distributed sensing network with improved detection accuracy introduces a high bandwidth overhead due to the frequent communication of spectrum data. In addition, many existing coherent spectrum sensing methods are not feasible for CBRS band due to the unknown operational characteristics of incumbent military wireless applications. To address these issues, we propose a machine learning based non-coherent spectrum sensing system: (F)eder(a)ted (I)ncumbent Detection in CB(R)S (FaIR). FaIR leverages a communication-efficient distributed learning framework, federated learning, for ESCs to collaborate and train a data-driven machine learning model for incumbent detection under minimal communication bandwidth. Our preliminary results show that the federated learning method can exploit the spatial diversity of ESCs and obtain an improved detection model comparing to a naive distributed sensing and centralized model framework. We evaluate the FaIR model with a variety of spectrum waveforms at varying SNRs. Our experiments showed that FaIR improves the average detection accuracy compared to the single-node method, using a fraction of the bandwidth compared to the naive multi-node method.
18
Communication-Efficient Learning of Deep Networks from Decentralized Data
H. Brendan McMahan, Eider Moore, Daniel Ramage et al. · arXiv (Cornell University) · 2016 · 5.2K citations · Full text
Federated Learning: Strategies for Improving Communication Efficiency
Jakub Konečný, H. Brendan McMahan, Felix X. Yu et al. · arXiv (Cornell University) · 2016 · 3K citations · Full text
Privacy-Preserving Deep Learning
Reza Shokri, Vitaly Shmatikov · 2015 · 2.3K citations