ResFed: Communication-Efficient Federated Learning With Deep Compressed Residuals
- Federated learning allows for cooperative training among distributed clients by sharing their locally learned model parameters, such as weights or gradients. However, as model size increases, the communication bandwidth required for deployment in wireless networks becomes a bottleneck. To address this, we propose a residual-based federated learning framework (ResFed) that transmits residuals instead of gradients or weights in networks. By predicting model updates at both clients and the server, residuals are calculated as the difference between updated and predicted models and contain more dense information than weights or gradients. We find that the residuals are less sensitive to an increasing compression ratio than other parameters, and hence use lossy compression techniques on residuals to improve communication efficiency for training in federated settings. With the same compression ratio, ResFed outperforms current methods (weight- or gradient-based federated learning) by over 1.4× on federated data sets, including MNIST,Federated learning allows for cooperative training among distributed clients by sharing their locally learned model parameters, such as weights or gradients. However, as model size increases, the communication bandwidth required for deployment in wireless networks becomes a bottleneck. To address this, we propose a residual-based federated learning framework (ResFed) that transmits residuals instead of gradients or weights in networks. By predicting model updates at both clients and the server, residuals are calculated as the difference between updated and predicted models and contain more dense information than weights or gradients. We find that the residuals are less sensitive to an increasing compression ratio than other parameters, and hence use lossy compression techniques on residuals to improve communication efficiency for training in federated settings. With the same compression ratio, ResFed outperforms current methods (weight- or gradient-based federated learning) by over 1.4× on federated data sets, including MNIST, FashionMNIST, SVHN, CIFAR-10, CIFAR-100, and FEMNIST, in client-to-server communication, and can also be applied to reduce communication costs for server-to-client communication.…
Author: | Rui SongORCiD, Liguo ZhouORCiD, Lingjuan Lyu, Andreas FestagORCiD, Alois KnollORCiD |
---|---|
Language: | English |
Document Type: | Article |
Year of first Publication: | 2023 |
published in (English): | IEEE Internet of Things Journal |
Publisher: | IEEE |
Place of publication: | New York |
ISSN: | 2327-4662 |
Volume: | 11 |
Issue: | 6 |
First Page: | 9458 |
Last Page: | 9472 |
Review: | peer-review |
Open Access: | ja |
Version: | published |
URN: | urn:nbn:de:bvb:573-46371 |
Related Identifier: | https://doi.org/10.1109/JIOT.2023.3324079 |
Faculties / Institutes / Organizations: | Fakultät Elektro- und Informationstechnik |
CARISSMA Institute of Electric, Connected and Secure Mobility (C-ECOS) | |
Fraunhofer-Anwendungszentrum "Vernetzte Mobilität und Infrastruktur" | |
Licence (German): | Creative Commons BY 4.0 |
Release Date: | 2024/03/25 |