HyperRS: Hypernetwork-Based Recommender System for the User Cold-Start Problem

Meta-learning has been proven to be effective for the cold-start problem of recommender systems. Many meta-learning systems that are designed user gradient-based. They use a global parameter learned from existing users initialize system new provides personalized recommendation with limited user-item interactions. require users’ demographic information learn parameter. This requirement raises privacy concerns, and is not always available. In addition, some gradient-based need dozens or even hundreds interactions difficult satisfy in specific scenarios which active their item rare. Moreover, rarely capture preferences over different attributes, as updating corresponding weights requires many optimization steps. We proposed HyperRS, Hypernetwork-based Recommender System, problem. Our does rely on provide recommendations. our system, hypernetwork generates all underlying system. The enables adapt quickly interest both attributes contents attributes. experimental results show method outperforms several state-of-the-art