Exploration-Exploitation Motivated Variational Auto-Encoder for Recommender Systems
This work addresses the need for recommender systems to recommend both familiar and novel items, which is an incremental improvement over existing collaborative filtering methods.
The paper tackles the problem of balancing known user preferences with novel item discovery in recommender systems by introducing XploVAE, a variational auto-encoder model that uses user-specific subgraphs for exploitation and exploration, resulting in demonstrated effectiveness on real-world datasets.
Recent years have witnessed rapid developments on collaborative filtering techniques for improving the performance of recommender systems due to the growing need of companies to help users discover new and relevant items. However, the majority of existing literature focuses on delivering items which match the user model learned from users' past preferences. A good recommendation model is expected to recommend items that are known to enjoy and items that are novel to try. In this work, we introduce an exploitation-exploration motivated variational auto-encoder (XploVAE) to collaborative filtering. To facilitate personalized recommendations, we construct user-specific subgraphs, which contain the first-order proximity capturing observed user-item interactions for exploitation and the high-order proximity for exploration. A hierarchical latent space model is utilized to learn the personalized item embedding for a given user, along with the population distribution of all user subgraphs. Finally, experimental results on various real-world datasets clearly demonstrate the effectiveness of our proposed model on leveraging the exploitation and exploration recommendation tasks.