Decentralized federated learning with privacy-preserving for recommendation systems

Hyperautomation can automate complex business processes, reduce human intervention and improve business operational efficiency. Recommendation systems (RS) facilitate hyperautomation greatly. However, these systems require a large amount of user data to train their machine learning (ML) models and hence user data privacy has received great attention. In this paper, we propose a decentralized federated learning framework with privacy-preserving for RS. In our framework, users train the private and public parameters locally but share the public parameters only. Extensive experiments verify that our approach is accurate and can well preserve privacy. This study is helpful for providing privacy preserving in hyperautomation.