FL2: Fuzzy Logic for Device Selection in Federated Learning

Federated Learning is an emerging paradigm in the field of Machine Learning which deals with distributed computing. Massively and geographically distributed clients participate in the training without using their private data and only send their locally trained model to the server. Since the clients may have different computational power, sizes of training sample labels with different distributions, it’s important to give different weights to each client based on these attributes. In this study, we proposed a method for giving different weights to clients using Fuzzy Logic. Results showed that the proposed approach performed better than the original Federated Averaging algorithm.