Scholarlink

We have explored the application of attention mechanisms within the federated learning framework, particularly in enhancing the performance for the specific task of predicting whether customers will click on advertisements. To conduct this research, we selected several widely used federated learning models, including FedAvg, FedProx, FedMA, and SCAFFOLD, and combined them with attention mechanisms. Our goal was to assess whether this integration could improve the performance of the models when dealing with a Ad click prediction dataset from Kaggle. Our experimental results indicate that integrating attention mechanisms into FL models can significantly enhance the models' performance on click-through rate prediction tasks. This suggests that attention mechanisms can help various federated models better capture the key features that influence user click decisions.

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

Federated Learning Combined with Attention Applied to Ad Click Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Guang Hu,
  • Jiaqi Wu

摘要

We have explored the application of attention mechanisms within the federated learning framework, particularly in enhancing the performance for the specific task of predicting whether customers will click on advertisements. To conduct this research, we selected several widely used federated learning models, including FedAvg, FedProx, FedMA, and SCAFFOLD, and combined them with attention mechanisms. Our goal was to assess whether this integration could improve the performance of the models when dealing with a Ad click prediction dataset from Kaggle. Our experimental results indicate that integrating attention mechanisms into FL models can significantly enhance the models' performance on click-through rate prediction tasks. This suggests that attention mechanisms can help various federated models better capture the key features that influence user click decisions.