17篇FedDAT: An Approach for Foundation Model Finetuning in MultiModalHeterogeneous Federated Learning

17篇FedDAT: An Approach for Foundation Model Finetuning in MultiModalHeterogeneous Federated Learning

第一部分:论文收录信息

  • FedDAT: An Approach for Foundation Model Finetuning in Multi-Modal Heterogeneous Federated Learning

  • 作者
    Haokun Chen, Yao Zhang, Denis Krompass, Jindong Gu, Volker Tresp

  • 会议
    AAAI 2024(The Thirty-Eighth AAAI Conference on Artificial Intelligence)

第二部分:解决的问题

在多模态异构联邦学习场景下,如何高效、安全地微调大规模基础模型(Foundation Models)?

具体挑战包括:

  1. 隐私限制(要用联邦学习)

    • 多个客户端(如医疗、工业场景)无法集中共享数据

    • 只能通过联邦学习方式协作训练

      </