Scholarlink

In view of the non-independent and identically distributed (non-IID) problem and data time dynamics faced by financial data participating in federated learning (FL), and to explore the impact of client-side continual learning and server-side multi-dimensional Shapley value evaluation on the global model, we proposed a Federated Multi-dimensional Shapley value continual learning of financial data (FedMscf) algorithm. Experimental results show that using FedMscf, the global model performance is significantly improved, highlighting its effectiveness in solving local model bias and global model drift, thereby promoting the fairness of model aggregation weights and enhancing the memory of previous data.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-dimensional Shapley Value Federated Continual Learning Algorithm for Financial Data

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kangning Yin,
  • Zhen Ding,
  • Xinhui Ji,
  • Zhihua Dong,
  • Zhiguo Wang

摘要

In view of the non-independent and identically distributed (non-IID) problem and data time dynamics faced by financial data participating in federated learning (FL), and to explore the impact of client-side continual learning and server-side multi-dimensional Shapley value evaluation on the global model, we proposed a Federated Multi-dimensional Shapley value continual learning of financial data (FedMscf) algorithm. Experimental results show that using FedMscf, the global model performance is significantly improved, highlighting its effectiveness in solving local model bias and global model drift, thereby promoting the fairness of model aggregation weights and enhancing the memory of previous data.