A systematic review on federated learning system: a new paradigm to machine learning

Federated learning is a machine learning technique that permits clients to train the model at a local site in a collaborative manner. It builds a global shared model on the basis of updates of the local model without exchanging data among multiple devices. Federated learning was introduced in 2016 with the goal of enabling local training as well as distributed machine learning training at the edge node’s level. It plays a vital role in terms of preserving the privacy of data while training the machine learning model on multiple devices. However, the introduction of federated learning into real-world applications exposes certain challenges in the training process, which affect the overall efficacy and efficiency of the federated learning model in real-world scenarios. As a result, an increasing number of researchers are now focusing on tackling the issues of FL and exploring various efficient research approaches to overcome these current obstacles. This paper systematically provides a detailed overview of federated learning, covering its definition, the need behind its development, privacy concepts, characteristics, and brief knowledge regarding different system components of federated learning. Different open-source frameworks that are available and used for implementing and solving problems related to federated learning have also been addressed in this article. Beyond this, the taxonomy of federated learning systems and different architectures for the same have also been discussed. In this paper, a brief comparison of related concepts with federated learning and a comparison among existing and popular federated learning studies proposed in different articles in the area of federated learning have also been summarized. In addition to the above-stated information, this article also provides brief information and a summary of various application areas of federated learning. Lastly, this paper briefly addresses the different challenges and prospects of research that lead to progress in this field.