The Performance Analysis of Federated Learning Methods for IoT with Big Data

ML that creates a global framework by gathering knowledge from a number of different dispersed edge clients. FL allows on-device training, keeps client information in private, and updates the frameworks. FL approaches presume computational capabilities at each edge-device/client, which may not always be the case. Due to its inherent distributed architecture, FL has the features that make it a good fit for IoT networks. Though, there are a few obstacles that are specific to FL, the maximum significant is training across significantly heterogeneous data sets on IoT devices. Numerous new studies contended to minimize consequences of diversity but obviously, the efficacy of potential resolutions has concluded researched or enumerated. This is because the homogeneity of devices and designs in the complicated Internet of Things (IoT) networks ultimately affects the FL learning process and produces conventional FL inappropriate to be straightforwardly organized. This research aimed to provide an introduction to FL, as well as a thorough examination of the problem statements and developing issues, especially those that arise when attempting to implement FL in heterogeneous IoT contexts using ML and DL techniques.