Federated Learning

Now we are in an era of technology transformation in our everyday life, where data play a key role in the decision making and bringing the action into reality. These data are collected from many distributed sources. Another important concept in this process is machine learning (ML) and data analytics. Federated learning (FL) is the term coined by Google. It facilitated the distributed learning process and shared the results to the outcomes to the central entity instead of conducting the complete learning process at the centre. In the traditional machine learning approach, data are brought to the model. Whereas FL brings ML techniques to the data used at end devices. In a nutshell, FL, which is also known as distributed learning deals with both centralized and distributed devices. The central entity(server) selects the statistical model which is to be trained. Then, it sends the trained model to multiple decentralized devices or servers. These distributed nodes train the model locally with their own data. Finally, the central entity pools the results from the distributed nodes and prepares a common model without accessing the data from the other nodes.1,2