Optimal Adaptive Power Control for Over-The-Air Federated Edge Learning Under Fading Channels

Yu, X; Xiao, B; Ni, W; Wang, X

Optimal Adaptive Power Control for Over-The-Air Federated Edge Learning Under Fading Channels

Yu, X Xiao, B Ni, W

Wang, X

Permalink

Publisher:: IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Publication Type:: Journal Article
Citation:: IEEE Transactions on Communications, 2023, 71, (9), pp. 5199-5213
Issue Date:: 2023-09-01

Closed Access

	Filename	Description	Size
	Optimal_Adaptive_Power_Control_for_Over-The-Air_Federated_Edge_Learning_Under_Fading_Channels.pdf	Published version	1.41 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Yu, X
dc.contributor.author	Xiao, B
dc.contributor.author	Ni, W https://orcid.org/0000-0002-4933-594X
dc.contributor.author	Wang, X
dc.date.accessioned	2024-04-17T01:15:23Z
dc.date.available	2024-04-17T01:15:23Z
dc.date.issued	2023-09-01
dc.identifier.citation	IEEE Transactions on Communications, 2023, 71, (9), pp. 5199-5213
dc.identifier.issn	0090-6778
dc.identifier.issn	1558-0857
dc.identifier.uri	http://hdl.handle.net/10453/177994
dc.description.abstract	Channel fading can have a strong impact on the convergence of over-the-air federated edge learning (OTA-FEEL). This paper develops a new and optimal power control policy to minimize the optimality gap of OTA-FEEL under any independently and identically distributed fading. Specifically, we reveal that the optimal power control policy takes a structure where the variance of the effective channel from a device to the server should be minimized when its mean is given. Following this structure, a novel nested optimization algorithm is developed to iteratively minimize the variance using the Lagrange-dual method and then optimize the mean of the effective channel using one-dimensional search. A quasi-closed-form expression of the optimal power control policy is derived. It is shown that the optimal adaptive power control for OTA-FEEL performs an integration of the 'channel inverting' strategy and the opposite 'channel-proportional' strategy to balance the mean and variance of the effective channel. We also generalize the new policy when the channel statistics are unknown a-priori, and show that the optimal policy can be asymptotically approached over time. Simulations confirm the superiority of the policy to its existing alternatives in the convergence speed and learning accuracy of OTA-FEEL.
dc.language	English
dc.publisher	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
dc.relation.ispartof	IEEE Transactions on Communications
dc.relation.isbasedon	10.1109/TCOMM.2023.3283790
dc.rights	info:eu-repo/semantics/closedAccess
dc.subject	0804 Data Format, 0906 Electrical and Electronic Engineering, 1005 Communications Technologies
dc.subject.classification	4006 Communications engineering
dc.subject.classification	4009 Electronics, sensors and digital hardware
dc.subject.classification	4606 Distributed computing and systems software
dc.title	Optimal Adaptive Power Control for Over-The-Air Federated Edge Learning Under Fading Channels
dc.type	Journal Article
utslib.citation.volume	71
utslib.for	0804 Data Format
utslib.for	0906 Electrical and Electronic Engineering
utslib.for	1005 Communications Technologies
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Electrical and Data Engineering
utslib.copyright.status	closed_access	*
dc.date.updated	2024-04-17T01:15:21Z
pubs.issue	9
pubs.publication-status	Published
pubs.volume	71
utslib.citation.issue	9

Abstract:

Channel fading can have a strong impact on the convergence of over-the-air federated edge learning (OTA-FEEL). This paper develops a new and optimal power control policy to minimize the optimality gap of OTA-FEEL under any independently and identically distributed fading. Specifically, we reveal that the optimal power control policy takes a structure where the variance of the effective channel from a device to the server should be minimized when its mean is given. Following this structure, a novel nested optimization algorithm is developed to iteratively minimize the variance using the Lagrange-dual method and then optimize the mean of the effective channel using one-dimensional search. A quasi-closed-form expression of the optimal power control policy is derived. It is shown that the optimal adaptive power control for OTA-FEEL performs an integration of the 'channel inverting' strategy and the opposite 'channel-proportional' strategy to balance the mean and variance of the effective channel. We also generalize the new policy when the channel statistics are unknown a-priori, and show that the optimal policy can be asymptotically approached over time. Simulations confirm the superiority of the policy to its existing alternatives in the convergence speed and learning accuracy of OTA-FEEL.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/177994