Meta-Learning for Online Update of Recommender Systems

Online recommender systems should be always aligned with users' current interest to accurately suggest items that each user would like. Since usually evolves over time, the update strategy flexible quickly catch from continuously generated new user-item interactions. Existing strategies focus either on importance of interaction or learning rate for parameter, but such one-directional flexibility is insufficient adapt varying relationships between interactions and parameters. In this paper, we propose MeLON, a meta-learning based novel online supports two-directional flexibility. It featured an adaptive parameter-interaction pair inducing learn up-to-date interest. The procedure MeLON optimized following approach: it learns how generate optimal rates future updates. Specifically, first enriches meaning previous identifies role parameter interaction; then combines these two pieces information rate. Theoretical analysis extensive evaluation three real-world datasets validate effectiveness MeLON.