An Insight into Federated Learning: A Collaborative Approach for Machine Learning

Integration of interconnected Internet of Things devices in smart cities, healthcare generates vast amount of data that can be analyzed to gain valuable insights. Intelligent decision-making, personalized solutions can be significantly improved with the use of machine learning on this data. Centralized machine learning has various concerns related to data collection, storage, privacy, and security. Distributed on-site machine learning tackle limitations of centralized ML, however, make several assumptions unrealistic for end devices with moderate resource capabilities. In federated learning, a collaborative machine learning approach is used to train a global model excluding the need of raw data to move from the client. End devices train the model locally with local datasets and share the updates with the server to finally train the global model. In this paper, we provide an in-depth analysis of federated learning discussing its benefits, basic principle, algorithm, and room to enhance accuracy, privacy, communication, and collaboration. The paper further discusses the role of FL in numerous domains showing real-world uses of federated learning framework.