Core Concepts
The author introduces the Federated Causal Multi-label Feature Selection (FedCMFS) algorithm to address the challenges of multi-label feature selection in a federated setting, focusing on causal relationships and data privacy.
Abstract
The study explores the need for causal multi-label feature selection in a federated setting due to data privacy concerns. The FedCMFS algorithm is proposed with three subroutines to address relevant features, missed true features, and false relevant features. Experimental results show FedCMFS outperforms other algorithms across various metrics and datasets.
The study emphasizes the importance of feature selection in high-dimensional multi-label data and introduces the FedCMFS algorithm to tackle this issue. By considering causal relationships and preserving data privacy, FedCMFS shows promising results in experimental evaluations.
Key points include:
Introduction of FedCMFS for causal multi-label feature selection in a federated setting.
Addressing challenges of relevant features, missed true features, and false relevant features.
Experimental results showcasing the effectiveness of FedCMFS compared to other algorithms.
Stats
The extensive experiments on 8 datasets have shown that FedCMFS is effect for causal multi-label feature selection in federated setting.
Average precision values range from 0.5735 to 0.9452 across different datasets and client numbers.
Coverage values vary between 0.0346 and 0.5407 for different datasets and client numbers.