Loading [a11y]/accessibility-menu.js
Adaptive Modulation for Wireless Federated Edge Learning | IEEE Journals & Magazine | IEEE Xplore

Adaptive Modulation for Wireless Federated Edge Learning


Abstract:

Wireless federated edge learning (FEEL) has been recently proposed to support the mobile artificial intelligence (AI) applications. Instead of transmitting local data to ...Show More

Abstract:

Wireless federated edge learning (FEEL) has been recently proposed to support the mobile artificial intelligence (AI) applications. Instead of transmitting local data to the edge server, the local learning updates or model parameters are uploaded over the wireless channels to protect the data privacy and security. However, the unreliable wireless channel will cause random packet error, which eventually has a significant impact on both the convergence rate and the learning latency. To cope with this problem, we propose a novel adaptive modulation scheme to achieve the balance between learning latency and convergence rate caused by stochastic channel error. Different from the conventional one, the modulation scheme in wireless FEEL system can be adjusted according to devices’ computing power, channel conditions, and training data importance. To further improve the FEEL performance, a joint optimization algorithm of spectrum allocation and modulation scheme selection is formulated to maximize the learning efficiency. Finally, we conduct extensive experiments based on different datasets. Test results show that the proposed algorithm can improve the convergence speed of the model training as compared with the conventional adaptive modulation schemes.
Page(s): 1096 - 1109
Date of Publication: 27 April 2023

ISSN Information:

Funding Agency:


References

References is not available for this document.