The project aims to develop easily implementable, scalable and privacy preserving solutions for in-network data management that takes into account a) the resource heterogeneity and explores the entire gamut of tradeoffs between them, and b) the specific importance levels of the data used by them. This project will bring together tools from several disciplines—machine learning, algorithms, and networking—to address the project goals. We aim to build a new theoretical framework that builds on federated learning, transfer learning, and reinforcement learning in creating a decentralised learning technique. We envision to implement our framework and use it our existing testbeds in USA and Finland. The key application use case investigated in the project pertains to the environmental sensing. This joint three-year project by Yale University and the University of Helsinki is supported by the two Academy of Finland flagships, the Finnish Center for AI (FCAI) and the 6GENESIS.

2022

2024