A multi-player game for studying federated machine learning incentive schemes

• Main Author: Chan, Beng Chung
• Other Authors: Yu Han
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/147923
• Institution: Nanyang Technological University

Laws governing privacy would almost certainly be strengthened and made more stringent. In the long run, Federated Learning as a business adoption would be advantageous because concerns like privacy protection can be solved while also exploiting big data. As a result, a proper structure to help Federated Learning in business is needed to facilitate and maintain a stable and long-lasting Federation. This is vital for Federation Learning’s growth. FedGame which was previously developed is a useful method for analysing human participation behaviour in the context of various reward schemes. Users of the system will be able to participate in Federated Learning as a data owner(business) and make decisions about which Federation to join. Data on human decision then can be used and analysed to create an effective reward system such as linear, equal, individual, union and shapley that promotes engagement and commitment of high-quality data while also allowing the Federation to function. FedForum is then created as a follow-up to FedGame, with a web forum for users to discuss their game play, how they respond to various reward schemes, and tactics they used when playing. Users can use the forum to address the new features in the game, whether they are working or not, what they like and hate, what they would like to see in the future, and what they would like to see less of. FedGame is currently being developed in beta. Anyone who has registered for the testing has access to FedForum where they can discuss and provide information. Offer the developer feedback on what works and what doesn't.