Java Federated Learning Framework Architecture

Data is fuel for machine learning approaches, but in several domains, data are sensitive and demand more security and even no data transferring at all. Examples are medicine, military, or private business data. Federated learning is an approach for distributed machine learning with only local data usage restrictions. The main goal is to design and implement a framework of Federated Learning for Java (FL4J) with decentralized learning which works with only local data per node. The framework should accept not only neural networks as model representation but also should be able to compute custom learning algorithm concurrently. This work describes the general principles of federated learning and suggests the framework for such approach implementation.