GPUMF: A GPU-Enpowered Collaborative Filtering Algorithm through Matrix Factorization
Feng Li, Shucheng Zhang, Yunming Ye, Xishuang Han · 2015
Recommender system is a core component in many intelligent service systems. A good personalized recommender is an important service to users. Collaborative Filtering (CF), an effective approach to recommendation, has been widely used in many real-life systems. Matrix Factorization (MF) is an important approach to CF, because MF has flexibility in dealing with various data aspects and other application-specific requirements. However, the large computational burden required by MF poses a challenge of speeding up the MF process. In the past few years, Graphics Processing Unit (GPU) has evolved into a very flexible and powerful many-core processor. By transforming the traditional MF model, we can exploit the large-scale parallelization features of a massively multithreaded GPU. The results on various types of data show that the proposed algorithm can be well suited for the massively parallel GPU architecture.