Knowledge Distillation For Wireless Edge Learning

TL;DR

This paper introduces a knowledge distillation framework for wireless edge learning that improves frame error prediction in congested environments by combining local, federated, and cloud-based training methods, enhancing robustness and privacy.

Contribution

It proposes a novel framework using knowledge distillation and synthetic oversampling to outperform traditional federated and local training in wireless edge learning scenarios.

Findings

Knowledge distillation improves prediction accuracy.

Synthetic oversampling maintains privacy while enhancing performance.

Framework is robust against high frame error rates.

Abstract

In this paper, we propose a framework for predicting frame errors in the collaborative spectrally congested wireless environments of the DARPA Spectrum Collaboration Challenge (SC2) via a recently collected dataset. We employ distributed deep edge learning that is shared among edge nodes and a central cloud. Using this close-to-practice dataset, we find that widely used federated learning approaches, specially those that are privacy preserving, are worse than local training for a wide range of settings. We hence utilize the synthetic minority oversampling technique to maintain privacy via avoiding the transfer of local data to the cloud, and utilize knowledge distillation with an aim to benefit from high cloud computing and storage capabilities. The proposed framework achieves overall better performance than both local and federated training approaches, while being robust against catastrophic failures as well as challenging channel conditions that result in high frame error rates.