Federated Recommendation with Additive Personalization

With rising concerns about privacy, developing recommendation systems in a federated setting become a new paradigm to develop next-generation Internet service architecture. However, existing approaches are usually derived from a distributed recommendation framework with an additional mechanism for privacy protection, thus most of them fail to fully exploit personalization in the new context of federated recommendation settings. In this paper, we propose a novel approach called Federated Recommendation with Additive Personalization (FedRAP) to enhance recommendation by learning user embedding and the user's personal view of item embeddings. Specifically, the proposed additive personalization is to add a personalized item embedding to a sparse global item embedding aggregated from all users. Moreover, a curriculum learning mechanism has been applied for additive personalization on item embeddings by gradually increasing regularization weights to mitigate the performance degradation caused by large variances among client-specific item embeddings. A unified formulation has been proposed with a sparse regularization of global item embeddings for reducing communication overhead. Experimental results on four real-world recommendation datasets demonstrate the effectiveness of FedRAP.