Federated Learning with Non Independent and Identically Distributed Data

The aim of this article is to study the effect of different Federated Learning strategies with several non-IID (Independent and Identically Distributed) data distributions on the test accuracy of the studied neural network model and compare it with the centralized training. In this work, we focus on implementation of three Flower framework algorithms with four non-IID data skews to conduct classification of dumpers cases. The experiments show that the test accuracy in centralized training was up to 85 % while in decentralized case was up to 81%.