考虑竞争环境的联邦学习多维契约激励机制研究.

Aiming at the existing incentive mechanisms cannot satisfy the demand of implementation for federated learning（FL） in competitive environment, this paper proposed a multi-dimensional contract-based incentive mechanism for FL in competitive environment. Firstly, this paper introduced the competition intensity to measure the competition relationship between the data owners and the task publisher. Secondly, this paper designed a novel incentive form named as MM（monetary-the FL model） combined incentive, which combined monetary with usufruct of FL model. Thirdly, based on MM combined incentive, this paper constructed optimization problems of the multi-dimensional contract-based incentive mechanism design for FL in competitive environment for the task publisher’s profit maximization under complete information scenario and incomplete information scenario respectively. Finally, experimental results show that the multi-dimensional contract-based incentive mechanism based on MM combined incentive improves the willingness of data owners to participate in FL in competitive environment. [ABSTRACT FROM AUTHOR]