TY - JOUR
T1 - Machine Intelligence at the Edge with Learning Centric Power Allocation
AU - Wang, Shuai
AU - Wu, Yik Chung
AU - Xia, Minghua
AU - Wang, Rui
AU - Poor, H. Vincent
N1 - Funding Information:
Manuscript received June 8, 2019; revised November 9, 2019 and May 24, 2020; accepted July 12, 2020. Date of publication July 28, 2020; date of current version November 11, 2020. This work was supported in part by the National Natural Science Foundation of China under Grant 61771232 and Grant 61671488, in part by the Shenzhen Basic Research Project under Grant JCYJ20190809142403596, in part by the Natural Science Foundation of Guangdong Province under Grant 2019B1515130003, in part by the U.S. National Science Foundation under Grant CCF-0939370 and Grant CCF-1908308, in part by the Major Science and Technology Special Project of Guangdong Province under Grant 2018B010114001, and in part by the Fundamental Research Funds for the Central Universities under Grant 191gjc04. The associate editor coordinating the review of this article and approving it for publication was X. Cheng. (Corresponding author: Yik-Chung Wu.) Shuai Wang is with the Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen 518055, China, and also with the Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen 518055, China (e-mail: wangs3@sustech.edu.cn).
Publisher Copyright:
© 2002-2012 IEEE.
PY - 2020/11
Y1 - 2020/11
N2 - While machine-type communication (MTC) devices generate considerable amounts of data, they often cannot process the data due to limited energy and computational power. To empower MTC with intelligence, edge machine learning has been proposed. However, power allocation in this paradigm requires maximizing the learning performance instead of the communication throughput, for which the celebrated water-filling and max-min fairness algorithms become inefficient. To this end, this paper proposes learning centric power allocation (LCPA), which provides a new perspective on radio resource allocation in learning driven scenarios. By employing 1) an empirical classification error model that is supported by learning theory and 2) an uncertainty sampling method that accounts for different distributions at users, LCPA is formulated as a nonconvex nonsmooth optimization problem, and is solved using a majorization minimization (MM) framework. To get deeper insights into LCPA, asymptotic analysis shows that the transmit powers are inversely proportional to the channel gains, and scale exponentially with the learning parameters. This is in contrast to traditional power allocations where quality of wireless channels is the only consideration. Last but not least, a large-scale optimization algorithm termed mirror-prox LCPA is further proposed to enable LCPA in large-scale settings. Extensive numerical results demonstrate that the proposed LCPA algorithms outperform traditional power allocation algorithms, and the large-scale optimization algorithm reduces the computation time by orders of magnitude compared with MM-based LCPA but still achieves competing learning performance.
AB - While machine-type communication (MTC) devices generate considerable amounts of data, they often cannot process the data due to limited energy and computational power. To empower MTC with intelligence, edge machine learning has been proposed. However, power allocation in this paradigm requires maximizing the learning performance instead of the communication throughput, for which the celebrated water-filling and max-min fairness algorithms become inefficient. To this end, this paper proposes learning centric power allocation (LCPA), which provides a new perspective on radio resource allocation in learning driven scenarios. By employing 1) an empirical classification error model that is supported by learning theory and 2) an uncertainty sampling method that accounts for different distributions at users, LCPA is formulated as a nonconvex nonsmooth optimization problem, and is solved using a majorization minimization (MM) framework. To get deeper insights into LCPA, asymptotic analysis shows that the transmit powers are inversely proportional to the channel gains, and scale exponentially with the learning parameters. This is in contrast to traditional power allocations where quality of wireless channels is the only consideration. Last but not least, a large-scale optimization algorithm termed mirror-prox LCPA is further proposed to enable LCPA in large-scale settings. Extensive numerical results demonstrate that the proposed LCPA algorithms outperform traditional power allocation algorithms, and the large-scale optimization algorithm reduces the computation time by orders of magnitude compared with MM-based LCPA but still achieves competing learning performance.
KW - Empirical classification error model
KW - edge machine learning
KW - learning centric communication
KW - multiple-input multiple-output
KW - resource allocation
UR - http://www.scopus.com/inward/record.url?scp=85095552871&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85095552871&partnerID=8YFLogxK
U2 - 10.1109/TWC.2020.3010522
DO - 10.1109/TWC.2020.3010522
M3 - Article
AN - SCOPUS:85095552871
SN - 1536-1276
VL - 19
SP - 7293
EP - 7308
JO - IEEE Transactions on Wireless Communications
JF - IEEE Transactions on Wireless Communications
IS - 11
M1 - 9151375
ER -