I. Introduction
In a broad sense, the recommendation system is essentially an information filtering system, where its core focus revolves around information that serves as a vital link connecting information producers and consumers. Conventional recommendation methods, including collaborative filtering and content-based filtering, have shown their effectiveness in capturing user preferences to make accurate recommendations. Nevertheless, these methods often face challenges in either handling sparse and high-dimensional data or capturing complex user-item interactions [1], [2]. To handle this, some scholars have applied autoencoders (e.g., V AE: variational autoencoder) to ranking recommendation systems, which alleviates the challenging issue to some extent [3], [8].