Federated low-rank tensor projections for sequential recommendation

作者：

Highlights：

• We propose a FedSeqRec framework, where several organizations can effectively collaborate to train an intelligent recommendation model.

• We introduce a federated re-average algorithm based on the similarity between the client data distribution and the overall data distribution.

• We propose an algebra-based sequential recommendation model that utilizes low-rank tensor projection to model the users’ dynamic preferences.

• Empirical results demonstrate that FedSeqRec outperforms state-of-the-art federated recommendation methods.

摘要

•We propose a FedSeqRec framework, where several organizations can effectively collaborate to train an intelligent recommendation model.•We introduce a federated re-average algorithm based on the similarity between the client data distribution and the overall data distribution.•We propose an algebra-based sequential recommendation model that utilizes low-rank tensor projection to model the users’ dynamic preferences.•Empirical results demonstrate that FedSeqRec outperforms state-of-the-art federated recommendation methods.

论文关键词：Federated learning,Federated recommendation,Sequential recommendation

论文评审过程：Received 22 July 2021, Revised 22 June 2022, Accepted 15 July 2022, Available online 22 July 2022, Version of Record 12 September 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109483