Summary
International Conference on Emerging Technologies for Communications
2022
Session Number:S8
Session:
Number:S8-2
Proposal for Auction Mechanism of Federated Learning with Knowledge Distillation
Yutaka Hatazawa, Takuji Tachibana,
pp.-
Publication Date:2022/11/29
Online ISSN:2188-5079
Summary:
In federated learning, knowledge distillation is useful to compress a learning model while maintaining learning accuracy. In this paper, we propose an auction mechanism of federated learning with knowledge distillation. In this mechanism, the utilization of knowledge distillation can be determined so that a social welfare is maximized. Numerical examples show that the proposed method can appropriately utilize the knowledge distillation to maximize social welfare.