Title: Aggregation techniques in wireless communication using federated learning: a survey

Authors: Gaganbir Kaur; Surender K. Grewal

Addresses: Department of Electronics & Communication Engineering, Deenbandhu Chhotu Ram University of Science and Technology (DCRUST), Murthal, Sonepat, Haryana, India ' Department of Electronics & Communication Engineering, Deenbandhu Chhotu Ram University of Science and Technology (DCRUST), Murthal, Sonepat, Haryana, India

Abstract: With the recent explosive rise in mobiles, IoT devices and smart gadgets, the data generated by these devices has grown exponentially. Given that the data generated by these devices is private, transmitting large amounts of private data is not practical. So, a new learning paradigm has been introduced known as federated learning, which is a machine learning technique. In this technique, user data is not transmitted to the base server as in centralised approach but only the locally updated model is transmitted. These model updates generated by the devices are aggregated at the server which updates its global model according to the local models and transmits back to the devices for next round. This technique reduces the privacy risk and also decreases the communication overhead. Various aggregation schemes are proposed in the literature for increasing the performance and accuracy of the system while also increasing the security and reliability. This paper presents a survey of the latest advances in research of such aggregation techniques.

Keywords: federated learning; machine learning; stochastic gradient descent; aggregation techniques; federated averaging; data privacy; wireless communication.

DOI: 10.1504/IJWMC.2024.137135

International Journal of Wireless and Mobile Computing, 2024 Vol.26 No.2, pp.115 - 126

Accepted: 19 Jul 2022
Published online: 04 Mar 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article