Rethinking Blockchain and Decentralized Learning: Position Paper

Blockchain technology and decentralized learning are attracting growing attention. Most existing methods of machine learning is in the centralized form which relies upon the third party in terms of the raw datasets and mining resources. Blockchain solves world centralization problems that keep the system secure through complex mathematical computations puzzle solved by blockchain miners. Concurrently, decentralized learning such as federated model allows the user to collaboratively access the updated prediction model without revealing the training data to the public. By doing so, it provides less power consumption, lower latency that respects the user’s privacy concern. Therefore, we study the extent to which these two technologies can be applied in the real world for faster convergence without compromising user’s security.