Please use this identifier to cite or link to this item: https://ruomo.lib.uom.gr/handle/7000/1714
Title: Federated Learning Approach Decouples Clients From Training a Local Model and With the Communication With the Server
Authors: Stergiou, Konstantinos D.
Psannis, Kostas E.
Type: Article
Subjects: FRASCATI::Engineering and technology
FRASCATI::Natural sciences
Keywords: Federated learning
mobile edge computing
traffic sign image recognition
convolutional neural networks
Issue Date: Dec-2022
Source: IEEE Transactions on Network and Service Management
Volume: 19
Issue: 4
First Page: 4213
Last Page: 4218
Abstract: Traffic sign recognition and autonomous vehicles computing are a few of the innovative applications which are emerging in the domain of mobile edge computing. Distributed machine learning in the form of Federated Learning (FL) has been applied to mobile edge computing through a range of methodologies and techniques for intelligent feature classification approaches. The challenges that research on such FL methods is facing is twofold: identify an optimal distributed architecture and algorithm components to each side to meet the demand of heavy data processing, and enhance the algorithm components with heuristics that fit to the problem domain and optimize the key parameters of the algorithms. In this prospect, we present a Federated Learning implementation based on a neural network architecture with emphasis to traffic sign image recognition. Our benchmark was tested with two FL strategies seeking an optimal performance model and in reference to a corresponding data set. We present the results of this work while we define the scope of future improvements to our model.
URI: https://doi.org/10.1109/TNSM.2022.3197059
https://ruomo.lib.uom.gr/handle/7000/1714
ISSN: 1932-4537
2373-7379
Other Identifiers: 10.1109/TNSM.2022.3197059
Appears in Collections:Department of Applied Informatics

Files in This Item:
File Description SizeFormat 
TNSM-2022-05053_final.pdf359,79 kBAdobe PDFView/Open


This item is licensed under a Creative Commons License Creative Commons