An Efficient Heterogeneous Edge-Cloud Learning Framework For Spectrum Data Compression

Guangyu Wu,Fuhui Zhou,Guoru Ding,Qihui Wu,Xiang-Yang Li

DOI: https://doi.org/10.1109/tmc.2022.3153049

IF: 6.075

2022-01-01

IEEE Transactions on Mobile Computing

Abstract:Spectrum data compression with a high-rate compression and accurate reconstruction is of crucial importance for reducing the ultra-large data transmission from the edge sensors to the cloud for establishing high-quality spectrum maps. However, the current methods ignore the imbalanced edge-cloud computation resources and cannot tackle the outlier signals, resulting in significant challenges for achieving effective compression. Therefore, we develop an efficient heterogeneous edge-cloud learning framework. In the framework, paralleled methods compress normal data and outlier data distinctively based on their different structure information. Meanwhile, those methods are asymmetric for achieving low-cost compression at the edge and accurate reconstruction on the cloud. Based on the framework, we propose an outlier-processable attention-based asymmetric compression algorithm. A novel attention-based asymmetric convolutional neural network performs the normal data compression while a non-linear outlier compression algorithm realizes the outlier data compression. Compared with the state-of-the-art schemes in real-world settings, our proposed framework’s convergence speed increases by 120% . Meanwhile, our framework’s reconstruction accuracy increases by 68.42% under the interfered environments while maintaining superior compression speed and comprehensive performance. We also confirm our framework’s generalization ability to transfer among different tasks by deploying it under various spectrum environments.

computer science, information systems,telecommunications