Full Text:
<14>
CLC number:
On-line Access: 2025-02-21
Received: 2024-06-20
Revision Accepted: 2024-12-15
Crosschecked: 0000-00-00
Cited: 0
Clicked: 20
FedMcon: an adaptive aggregation method for federated learning via meta controller
| |||||||||||||
Tao SHEN1, Zexi LI1, Ziyu ZHAO1, Didi ZHU1, Zheqi LV1, Kun KUANG1, Shengyu ZHANG2‡, Chao WU3,4‡, Fei WU1‡. FedMcon: an adaptive aggregation method for federated learning via meta controller[J]. Frontiers of Information Technology & Electronic Engineering, 1998, -1(-1): .
@article{title="FedMcon: an adaptive aggregation method for federated learning via meta controller",
author="Tao SHEN1, Zexi LI1, Ziyu ZHAO1, Didi ZHU1, Zheqi LV1, Kun KUANG1, Shengyu ZHANG2‡, Chao WU3,4‡, Fei WU1‡",
journal="Frontiers of Information Technology & Electronic Engineering",
volume="-1",
number="-1",
pages="",
year="1998",
publisher="Zhejiang University Press & Springer",
doi="10.1631/FITEE.2400530"
}
%0 Journal Article
%T FedMcon: an adaptive aggregation method for federated learning via meta controller
%A Tao SHEN1
%A Zexi LI1
%A Ziyu ZHAO1
%A Didi ZHU1
%A Zheqi LV1
%A Kun KUANG1
%A Shengyu ZHANG2‡
%A Chao WU3
%A 4‡
%A Fei WU1‡
%J Journal of Zhejiang University SCIENCE C
%V -1
%N -1
%P
%@ 2095-9184
%D 1998
%I Zhejiang University Press & Springer
%DOI 10.1631/FITEE.2400530
TY - JOUR
T1 - FedMcon: an adaptive aggregation method for federated learning via meta controller
A1 - Tao SHEN1
A1 - Zexi LI1
A1 - Ziyu ZHAO1
A1 - Didi ZHU1
A1 - Zheqi LV1
A1 - Kun KUANG1
A1 - Shengyu ZHANG2‡
A1 - Chao WU3
A1 - 4‡
A1 - Fei WU1‡
J0 - Journal of Zhejiang University Science C
VL - -1
IS - -1
SP -
EP -
%@ 2095-9184
Y1 - 1998
PB - Zhejiang University Press & Springer
ER -
DOI - 10.1631/FITEE.2400530
Abstract: federated learning (FL) emerged as a novel machine learning setting that enables collaboratively training deep models on decentralized clients with privacy constraints. In vanilla federated learning algorithm (FEDAVG), the global model is generated by the weighted linear combination of local models, and the weights are proportional to the local data sizes. This methodology, however, encounters challenges when facing heterogeneous and unknown client data distributions, often leading to discrepancies from the intended global objective. The linear-combination-based aggregation often fails to address the varied dynamics presented by diverse scenarios, settings, and data distributions inherent in FL, resulting in hindered convergence and compromised generalization. In this paper, we presents a new aggregation method, FedMcon, in a framework of meta learning for FL. We introduce a learnable controller that trained on a small proxy dataset and served as the aggregator to learn how to adaptively aggregate heterogeneous local models into a better global model toward the desired objective. The experimental results indicate that the proposed method is effective in extremely non-i.i.d. data and it can simultaneously reach 8.5% generalization gain and 19 times communication speedup in one FL setting.
Open peer comments: Debate/Discuss/Question/Opinion
<1>