Self-attention fusion and adaptive continual updating for multimodal federated learning with heterogeneous data.

Journal: Neural networks : the official journal of the International Neural Network Society
PMID:

Abstract

Federated learning (FL) enables collaborative model training without direct data sharing, facilitating knowledge exchange while ensuring data privacy. Multimodal federated learning (MFL) is particularly advantageous for decentralized multimodal data, effectively managing heterogeneous information across modalities. However, the diversity in environments and data collection methods among participating devices introduces substantial challenges due to non-independent and identically distributed (non-IID) data. Our experiments reveal that, despite the theoretical benefits of multimodal data, MFL under non-IID conditions often exhibits poor performance, even trailing traditional unimodal FL approaches. Additionally, MFL frequently encounter missing modality issues, further complicating the training process. To address these challenges, we propose several improvements: the federated self-attention multimodal (FSM) feature fusion method and the multimodal federated learning adaptive continual update (FedMAC) algorithm. Moreover, we utilize a Stable Diffusion model to mitigate the impact of missing image modality. Extensive experimental results demonstrate that our proposed methods outperform other state-of-the-art FL algorithms, enhancing both accuracy and robustness in MFL.

Authors

  • Kangning Yin
    School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu, 610054, PR China; Institute of Public Security, Kash Institute of Electronics and Information Industry, Kashi, Xinjiang, 84400, PR China. Electronic address: knyin@std.uestc.edu.cn.
  • Zhen Ding
    School of Public Health, Xuzhou Medical College, Xuzhou, 209 Tong-Shan Road, Xuzhou, Jiangsu, 221002, China.
  • Xinhui Ji
    Shanghai New Energy Vehicle Public Data Collection and Monitoring Research Center, Shanghai, 201800, PR China. Electronic address: jixinhui@shevdc.org.
  • Zhiguo Wang
    Sino-Dutch Biomedical and Information Engineering School, Northeastern University, Shenyang, 110169, Liaoning, China.