Deep Learning-Based Retrieval Algorithms for Recommendation Systems
Suling Zhao, Zhen Huang, Huangmei Guo, Xuan Xiao, Renzhen Ye · 2023
The recommendation system is an important technology to solve information overload and capture users' interests, which is widely used in recommendation platforms such as e-commerce, movies, and books. The overall process of a recommendation system is that candidate items enter the recommendation system, a small number of items that users may be interested in are filtered by the retrieval layer, and then the final recommendation list is obtained by more precise filtering by the ranking layer. As the first stage of the recommendation system, the retrieval layer’s efficiency determines the ability of the recommendation system to quickly filter information, and it determines the upper limit of the ranking effect. However, few works have elaborated the retrieval layer in detail. Since the great achievements of deep learning techniques in recommendation systems, this paper focuses on the deep learning methods often used in the retrieval layer, which is of referenceable significance for the optimization of recommendation systems.