Designing Incentive Controls for Federated Learning in Aviation
Implemented on-chain
B.R. Lapré (TU Delft - Technology, Policy and Management)
G.A. de Reuver – Graduation committee member (TU Delft - Technology, Policy and Management)
Marcela Tuler de Oliveira – Graduation committee member (TU Delft - Technology, Policy and Management)
R. van Bergem – Graduation committee member (TU Delft - Technology, Policy and Management)
A. Apostelodis – Graduation committee member (KLM Royal Dutch Airlines)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Federated learning is a promising method of distributed machine learning that can allow industries such as aviation to utilize their data without actively sharing it, but has issues regarding cooperation if proper incentives are not utilized. This thesis explores the problem and builds from a previously designed FL Framework by designing an IT system that can be used to control FL Tasks and Rewards. Using an institutional analysis to set the playing field and based on previous work, an IT artefact consisting of a set of smart contracts on a private blockchain were designed and implemented as an MVP. These contracts perform the role of organising and controlling the participants within FL tasks and can be used to distribute rewards in the form of two types of Tokens: The Reward Token which also can be utilized for voting in proposals and reward distributing as a representation of ownership, and the Model Access Token, which can be used to represent Model access.