Frequency-aware Graph Signal Processing for Collaborative Filtering

Graph Signal Processing (GSP) based recommendation algorithms have recently attracted lots of attention for high efficiency. However, these methods failed to utilize user/item unique characteristics, as well as user and item high-order neighborhood information when modeling user preference, leading to sub-optimal performance. To this end, we propose a frequency-aware graph signal processing method (FaGSP) for collaborative filtering (CF). Firstly, we design a Cascaded Filter Module to capture both unique and common user/item characteristics to more accurately model user preference. Then, we devise a Parallel Filter Module to fully utilize high-order neighborhood information of users/items for more accurate user preference modeling. Finally, we combine these modules via a linear model to further improve recommendation accuracy. Extensive experiments demonstrate the superiority of our method from prediction accuracy and training efficiency.